Publication number  US7617119 B1 
Publication type  Grant 
Application number  US 09/888,340 
Publication date  Nov 10, 2009 
Priority date  Dec 20, 2000 
Fee status  Paid 
Publication number  09888340, 888340, US 7617119 B1, US 7617119B1, USB17617119, US7617119 B1, US7617119B1 
Inventors  Michael Neal, Krishna Venkatraman, Rob Parkin, Suzanne Valentine, Phil Delurgio, Hau Lee 
Original Assignee  Demandtec, Inc. 
Export Citation  BiBTeX, EndNote, RefMan 
Patent Citations (107), NonPatent Citations (88), Referenced by (44), Classifications (13), Legal Events (6)  
External Links: USPTO, USPTO Assignment, Espacenet  
This application is a Continuationin Part to application Ser. No. 09/741,958, filed Dec. 20, 2000, now U.S. Pat. No. 7,523,047 entitled “Price Optimization System”, by Michael Neal, Krishna Venkatraman, Suzanne Valentine, Phil Delurgio, and Hau Lee, and application Ser. No. 09/742,472, filed Dec. 20, 2000, entitled “Imputed Variable Generator”, by Suzanne Valentine, Krishna Venkatraman, Phil Delurgio, Michael Neal, and Hau Lee, and copending application Ser. No. 09/741,956, filed Dec. 20, 2000, entitled “Econometric Engine”, by Hau Lee, Krishna Venkatraman, Michael Neal, Suzanne Valentine, and Phil Delurgio, and copending application Ser. No. 09/741,957, filed Dec. 20, 2000, entitled “Financial Model Engine”, by Phil Delurgio, Suzanne Valentine, Michael Neal, Krishna Venkatraman, and Hau Lee, and copending application Ser. No. 09/741,959, filed Dec. 20, 2000, entitled “Econometric Optimization Engine”, by Krishna Venkatraman, Phil Delurgio, Suzanne Valentine, Michael Neal, and Hau Lee, and copending application Ser. No. 09/849,783, filed May 5, 2001, entitled “System For Creating Optimized Promotion Event Calendar”, by Michael Neal, Krishna Venkatraman, Rob Parkin, Suzanne Valentine, Phil Delurgio, and Hau Lee which are all incorporated by reference herein for all purposes.
This application also relates to copending application Ser. No. 09/849,621, filed May 5, 2001, entitled “Apparatus For Merchandise Promotion Optimization”, by Michael Neal, Krishna Venkatraman, Suzanne Valentine, Phil Delurgio, and Hau Lee,
The present invention relates to providing optimized pricing for a plurality of products for a plurality of stores.
In businesses, prices of various products must be set. Such prices may be set with the goal of maximizing profit or demand or for a variety of other objectives. Profit is the difference between total revenue and costs. Total sales revenue is a function of demand and price, where demand is a function of price. Demand may also depend on the day of the week, the time of the year, the price of related products, location of a store, and various other factors. As a result, the function for forecasting demand may be very complex. Costs may be fixed or variable and may be dependent on demand. As a result, the function for forecasting costs may be very complex. For a chain of stores with tens of thousands of different products, forecasting costs and determining a function for forecasting demand are difficult. The enormous amounts of data that must be processed for such determinations are too cumbersome even when done by computer. Further, the methodologies used to forecast demand and cost require the utilization of nonobvious, highly sophisticated statistical processes.
It is desirable to provide an efficient process and methodology for determining the prices of individual products such that profit (or whatever alternative objective) is optimized.
In some of the above mentioned patent applications, a plurality of rules are used as constraints in providing an optimization. During optimization it may be found that one of the plurality of rules is infeasible. Such a finding of infeasibility may stop the optimization process. It would be desirable to provide a method of handling optimization when a rule is found to be infeasible.
To achieve the foregoing and other objects and in accordance with the purpose of the present invention, a method for computing a preferred set of prices for a plurality of products, comprising generating a preferred set of prices, is provided. The generating a preferred set of prices, comprises, storing a plurality of rules, allowing the prioritization of the plurality of rules, and relaxing at least one lower priority rule to allow higher priority rules to become feasible.
In another embodiment of the invention, an apparatus for computing a preferred set of prices for a plurality of products is provided. An optimization engine is provided, which comprises computer readable media, which comprises computer readable code for storing a plurality of rules, computer readable code for allowing the prioritization of the plurality of rules, and computer readable code for relaxing at least one lower priority rule to allow higher priority rules to become feasible.
These and other features of the present invention will be described in more detail below in the detailed description of the invention and in conjunction with the following figures.
The present invention is illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings and in which like reference numerals refer to similar elements and in which:
The present invention will now be described in detail with reference to a few preferred embodiments thereof as illustrated in the accompanying drawings. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. It will be apparent, however, to one skilled in the art, that the present invention may be practiced without some or all of these specific details. In other instances, well known process steps and/or structures have not been described in detail in order to not unnecessarily obscure the present invention.
I. Overall System
To facilitate discussion,
Each component of the price optimizing system 100 will be discussed separately in more detail below.
II. Econometric Engine
A. Imputed Variable Generator
The present invention provides methods, media, and systems for generating a plurality of imputed econometric variables. Such variables are useful in that they aid businesses in determining the effectiveness of a variety of sales strategies. In particular, such variables can be used to gauge the effects of various pricing or sales volume strategies.
1. Initial Dataset Creation and Cleaning
The process of dataset creation and cleaning (that is to say the process of identifying incompatible data records and resolving the data incompatibility, also referred to herein as “error detection and correction”) begins by inputting raw econometric data (Step 1011). The raw econometric data is then subject to formatting and classifying by UPC designation (Step 1013). After formatting, the data is subject an initial error detection and correction step (Step 1015). Once the econometric data has been corrected, the store information comprising part of the raw econometric data is used in defining a store data set hierarchy (Step 1017). This is followed by a second error detecting and correcting step (Step 1019). This is followed by defining a group of products which will comprise a demand group (i.e., a group of highly substitutable products) and be used for generating attribute information (Step 1021). Based on the defined demand group, the attribute information is updated (Step 1023). The data is equivalized and the demand group is further classified in accordance with size parameters (Step 1025). The demand group information is subjected to a third error detection and correction step (Step 1027). The demand group information is then manipulated to facilitate decreased process time (Step 1029). The data is then subjected to a fourth error detection and correction step (Step 1031), which generates an initial cleansed dataset. Using this initial cleansed dataset, imputed econometric variables are generated (Step 1033). Optionally, these imputed econometric variables may be output to other systems for further processing and analysis (Step 1035).
The process begins by inputting raw econometric data (Step 1011). The raw econometric data is provided by a client. The raw econometric data includes a variety of product information. The raw econometric data must specify the store from which the data is collected, the time period over which the data is collected and include a UPC (Universal Product Code) for the product, and provide a UPC description of the product. Also, the raw econometric data must include product cost (e.g., the wholesale cost to the store), number of units sold, and either unit revenue or unit price. Ordinarily, the UPC description also identifies the product brand, UOM (Unit of Measure), and product size. Such information can be very detailed or less so. For example, brand can simply be CocaCola®, or more detailed e.g., Cherry CocaCola®. A UOM is, for example, ounces (oz.), pound (lb.), liter (ltr), or count (CT). Size reflects the number of UOM's e.g., eight (8) oz or two (2) ltr. Also, the general category of product or department identification is input. A category is defined as a set of substitutable or complementary products, for example, “Italian Foods”. Such categorization can be proscribed by the client, or defined by generally accepted product categories. Additionally, such categorization can be accomplished using lookup tables or computer generated product categories.
Also, a more complete product descriptor is generated using the product information described above and, for example, a UPC description of the product and/or a product description found in some other lookup table (Step 1013). This information is incorporated into a product format. This product format provides a more complete picture of the product, but this information is stored in a separate database which is not necessarily processed using the invention. This information provides a detailed description of the product which can be called up as needed.
The data is then subjected to a first error detection and correction process (Step 1015). Typically, this step includes the removal of all duplicate records and the removal of all records having no match in the client supplied data (typically scanner data). An example of records having no match are records that appear for products that the client does not carry or stock in its stores. These records are detected and deleted.
Data subsets concerning store hierarchy are defined (Step 1017). This means stores are identified and categorized into various useful subsets. Typical subsets include (among other categorizations) stores segregated by, for example, zip codes, cities, states, specific geographical regions, rural environs, urban environs, associations with other stores (e.g., is this store part of a mall) or categorized by specific stores. A wide variety of other subsets may also be used. These subsets can be used to provide information concerning, among other things, regional or location specific economic effects.
The data is then subjected to a second error detection and correction process (Step 1019). This step cleans out certain obviously defective records. Examples include, but are not limited to, records displaying negative prices, negative sales volume, or negative cost. Records exhibiting unusual price information are also removed. Such unusual prices can be detected using cross store price comparisons (between similarly situated stores), for example, an average price for a product in the stores of a particular geographic region can be determined by averaging the prices for all such products sold in the subject stores. The standard deviation can also be calculated. Prices that lie at greater than, for example, two (2) standard deviations from the mean price will be treated as erroneous and such records will be deleted. These tools can be applied to a variety of product parameters (e.g., price, cost, sales volume).
This is followed by defining groups of products and their attributes and exporting this information to a supplementary file (e.g., a text file) (Step 1021). This product information can then be output into a separate process which can be used to define demand groups or product attributes. For example, this supplemental file can be input into a spreadsheet program (e.g., Excel®) which can use the product information to define “demand groups” (i.e. groups of highly substitutable products). Also, further product attribute information can be acquired and added to the supplementary file. Such attributes can comprise, for example, branding information, manufacturer, size, flavor or form (e.g., cherry soda) just to name a few. Such information can be gleaned from multiple sources e.g., UPC product catalogues, the client, product lookup tables, or other sources. The advantage of such supplementary files is that they maintain complete product information (including information not required by the processes of the present invention) which can be accessed when needed. In addition, updated demand group and attribute information can then be input as received (Step 1023). By maintaining a supplementary file containing large amounts of data, a more streamlined (abbreviated) dataset may be used in processing. This effectively speeds up processing time by deleting noncritical information from the dataset.
The data is further processed by defining an “equivalizing factor” for the products of each demand group in accordance with size and UOM parameters (Step 1025). This equivalizing factor can be provided by the client or imputed. An example of determining an imputed equivalizing factor follows. Product size and UOM information are obtained, for example, from the product description information. Typical examples of such size and UOM information is, 20 oz. (ounce), 6 CT (count), or 1 ltr (liter). A further advantageous aspect of the present invention is that, even if such size or UOM information is incomplete or not provided, it can also be imputed. An equivalizing factor can be imputed by using, for example, the median size for each UOM. Alternatively, some commonly used arbitrary value can be assigned. Once this information is gathered, all product prices and volume can be “equivalized”. In one example, a demand group (a group of highly substitutable products) is chosen having, for example, “soft drinks” as its subject category. And by further example, the soft drink product comes in 8, 12, 16, 24, 32, and 64 ounce sizes. The median size (or for that matter, any arbitrarily determined size) can then be used as the base size to which all other sizes are to be equivalized. For example, using the 8, 12, 16, 24, 32, and 64ounce sizes discussed above, an arbitrary base size can be determined as, for example, 24 ounces. Then the 24ounce size is determined as the equivalizing factor. Some of the uses of the equivalizing factors are detailed in the discussions below. Chiefly, the purpose of determining an equivalizing factor is to facilitate comparisons between different size products in a demand group. For example, if 16 is determined as the equivalizing factor for the above group of soft drinks, then an 8 oz. soft drink is equivalized to one half of a 16 oz. unit. In a related vein, a 32 oz. soft drink is equivalized to two (2) 16 oz. units.
Additionally, size information can be used to define further product attributes. For example, if the size is in the bottom tertile of sizes for that product, it will be classified as “Small” size. Correspondingly, if the size is in the middle tertile of sizes for that product, it will be classified as “Medium” size, and if the size is in the top tertile of sizes for that product, it will be classified as “Large” size. Such categorization can define product attributes such as small (8 and 12ounce sizes), medium (16 and 24ounce sizes), and large (32 and 64ounce sizes) sizes.
The data is then subjected to a third error detection and correction process, which detects the effects of closed stores and certain other erroneous records (Step 1027). Keeping in mind that one advantage of the present invention is that very little client input is required to achieve accurate results, the inventors contemplate error correction without further input (or very little input) from the client. In accord with the principles of the invention, stores that demonstrate no product movement (product sales equal to zero) over a predetermined time period are treated as closed. Those stores and their records are dropped from the process. In a preferred embodiment, the predetermined time period is three (3) months.
With continued reference to
Using the following illustrative table:
Record  Causal  
Number  Date  Store  Product  Units  Price  Variable 
1  12/5  Y  D  10  1.99  1 
2  12/5  Y  D  10  1.99  1 
3  12/12  Y  D  10  1.99  1 
4  12/12  Y  D  15  1.89  2 
5  12/19  Y  D  12  1.99  1 
6  12/26  Y  D  9  1.99  1 
Discrepant records are identified and corrected. For example, if two records have the same exact values (such as record #1 and record #2), it is assumed that one such record is an erroneous duplicate and only one record is kept as part of the analyzed dataset, for example, only record #1 is retained.
If two records with the same date, product id, and store id have multiple records with different causals, they are combined into a single record, with the two prices maintained in separate dataset variables, units summed across the two records, and the causal variables representing something other than a normal state being represented by new dataset variables.
The following table shows an updated version of the above table. Record 2 was deleted because it is identical to Record 1. Records 3 and 4 were combined into a single record (i.e., combined into a single Record 3) with new causal variables defined for Advertisement and Advertisement Price. Records 5 and 6 did not change because there was no duplicate information.
Record  Regular  Adver  Advertise  
Number  Date  Store  Product  Units  Price  tisement  ment Price 
1  12/5  Y  D  25  1.99  No  . 
3  12/12  Y  D  25  1.99  Yes  1.89 
5  12/19  Y  D  12  1.99  No  . 
6  12/26  Y  D  9  1.99  No  . 
A further correction can be made for records having the same date and causal value but have differing prices or differing number of units sold. First, a data discrepancy must be detected. For example, if a record on a specific date in the same store for the same product and causal state has two different values for units, this is a discrepancy. Correction can be accomplished by, first calculating the average number of units sold over all dates in the modeled time interval. The discrepant records are compared with the average value. The record having the unit value closest to the calculated average units is kept and the other is discarded. The same general process can be followed for records having discrepancies as to price (i.e., the record having the price closest to average price is kept). If both price and units are determined to have a discrepancy, the record having the price and unit values closest to the average price and average units is kept.
After all the duplicate records eliminated, the data is reconstructed. The data can be reviewed again to insure all duplicates are removed. Optionally, an output file including all discrepancies can be produced. In the event that it becomes necessary, this output file can be used as a followup record for consulting with the client to confirm the accuracy of the error detection and correction process.
Additionally, reduced processing times may be achieved by reformatting the data (Step 1029). For example, groups of related low sales volume products (frequently high priced items) can optionally be aggregated as a single product and processed together. Additionally, the data may be split into conveniently sized data subsets defined by a store or groups of stores which are then processed together to shorten the processing times. For example, all stores in the state of California can be processed together, then all the stores in Texas, etc.
Next the process includes determining the nature of missing data records in a fourth error detection and correction step (Step 1031). The missing data records are analyzed again before finally outputting a cleansed initial dataset. For example, data collected over a modeled time interval is analyzed by introducing the data into a data grid divided into a set of time periods. The time periods can be preset, computer determined, or user defined. The time periods can include, but are not limited to, months, weeks, days, or hours. One preferred embodiment uses time periods of one week. The data grid so constructed is analyzed. For the time periods (e.g., weeks) having no records a determination must be made. Is the record missing because:
a. there were no sales that product during that week (time period);
b. the product was sold out and no stock was present in the store during that time period (this situation is also referred to herein as a “stockout’);
c. the absence of data is due to a processing error.
Referring again to
Grid  Date  Store  Product  Units  Price 
1  1/7  Z  Y  10  1.99 
2  1/14  Z  Y  12  2.19 
3  
4  1/28  Z  Y  8  1.99 
5  2/7  Z  Y  10  1.99 
6  
Review of the grid (Step 1103) shows that records are “missing” for dates 1/21 and 2/14 (i.e., grid 3 and grid 6). Placeholders are set in the records defined by grid 3 and grid 6 (Step 11105). For example, an easily detectable or arbitrarily large value can be put in the price column of the grid, e.g. 999. Alternatively, a simple X can be placed as a placeholder in the price column. In the present example, “X's” can be placed in the price columns of grid 3 and grid 6.
If the first or last grid in the dataset (here grid 1 or grid 6) has few or no observations, those records are deleted from the dataset (Step 1107). For purposes of the above analysis, a grid having “few” observations is defined as a grid having 50% fewer observations than is normal for the grids in the middle of the dataset. Here, for example, the record for grid 6 (the last week) is deleted because no records are present for that week. Also, using clientsupplied stocking information, products which have been discontinued during the modeled time interval do not have their grids filled out for the discontinued time period (Step 1109). Also, products which are introduced during the modeled time interval have their time grid filled out only for those time periods occurring after the product introduction date. Thus, certain data aberrations are removed from the modeled dataset, permitting more accurate modeling.
The mean units (sold) and the STD for units are then calculated (Step 1111). For example, in dataset depicted above, the mean is 10 units. The missing record is then compared with the mean value (Step 1113). Here, a missing record (grid 3) is assigned an initial unit value=0. If the value of zero units lies within one (1) STD of the calculated mean, it is assumed that an actual value of zero units is feasible and that record is treated as if the record is valid (unit volume=0). However, if zero lies at greater than one STD from the mean, it is assumed that the value of zero units is due to a “stockout”. In such case, it is assumed that had product been present in the store an average number of units would have been sold. Therefore, the zero unit value for that record is replaced by a unit value equal to the calculated mean unit value, thereby correcting for the “stockout”. In this case, units for grid 3 will be corrected to calculated mean units (i.e., 10).
The product histories of the dataset can also be examined. If the subject product was introduced or discontinued as a salable item at the subject store during the modeled time interval, the grid is not filled out (with either zero or average values) for those time periods where the product was not offered for sale in the subject store. In this way missing records do not corrupt the dataset.
Further aspects of the fourth error detection and correction include a detection and elimination of outlying price data points (outliers). A satisfactory way of accomplishing this begins with a calculation of the mean price for each product within a given store, as determined over the modeled time interval. Once a mean price and STD are determined, all price data for the modeled time interval is examined. If it is determined that a price record lies within three (3) STD from the mean price it is deemed accurate and not an outlier. However, prices lying outside three (3) STD are treated as outliers. These outliers are assigned adjusted prices. The adjusted prices have the value of the immediately preceding time period (e.g., the previous day's or week's price for that product within the store). This adjusted price data is again checked for outliers (using the original mean and STD). Again, outliers are checked against the original mean and STD and again price adjusted if necessary. This usually removes all the remaining outliers. However, the process may optionally continue, iteratively, until there are no further outliers.
The net result of execution of the process Steps 10111031 disclosed hereinabove is the generation of a cleansed initial dataset which can be used for its own purpose or input into other econometric processes. One such process is the generation of imputed econometric variables.
2. Generation of Imputed Econometric Variables
The foregoing steps (10111031) concern cleansing the raw econometric data to create an error detected and error corrected (“cleansed”) initial dataset. The cleansed initial dataset created in the foregoing steps can now be used to generate a variety of useful imputed econometric variables (Step 1033). These imputed econometric variables are useful in their own right and may also be output for use in further processing (Step 1035). One particularly useful application of the imputed econometric variables is that they can be input into an optimization engine which collects data input from a variety of sources and processes the data to provide very accurate economic modeling information.
a. Imputed Base Price
One imputed econometric variable that can be determined using the initial dataset created in accordance with the forgoing, is an imputed base price variable (or base price).
Referring to
Referring again to
Alternatively, the initial base price can be determined using other methods. For example, the average price of the product over the −T/2 time period (4 weeks) preceding the data point X may be used as the initial base price. Whatever method used, the initial base price is generated for each time period of the modeled time interval. One by one, each data point in the modeled time frame is examined and an initial base price is determined for each time period (e.g., “X”) in the modeled time interval.
The initial base price values generated above provide satisfactory values for the imputed base price variable which may be output (Step 1207) and used for most purposes. However, optional Steps 12091217 describe an approach for generating a more refined imputed base price variable.
In generating a more refined imputed base price variable, the effect of promotional (or discount) pricing is addressed (Steps 12091217). This may be calculated by specifying a discount criteria (Step 1209); defining price steps (Step 1211); outputting an imputed base price variable and an imputed discount variable (Step 1213); analyzing the base price distribution (Step 1215); and outputting a refined base price variable (Step 1217).
Data records are evaluated over a series of time periods (e.g., weeks) and evaluated. The point is to identify price records which are discounted below a base price. By identifying these prices and not including them in a calculation of base price, the base price calculation will be more accurate. Therefore, a discount criterion is defined and input as a variable (Step 1209). A preferred criterion is 2%. Therefore, records having prices which are discounted 2% below the previously determined initial base price are treated as records having “promotional prices”. These records are temporarily deleted from the dataset. The remaining records, having zero or small discounts, are treated as “nonpromoted” records. So the price of each product for the “nonpromoted” time periods (weeks) is averaged over all time periods (weeks) in the modeled time interval. The average nonpromoted price is referred to as a base price.
Further analysis is used to define base price “steps” (Step 1211). This process can be more readily illustrated with references to
Further refining includes an analysis of the first step 1240. If the first step 1240 is short (along the time axis) and considerably lower than the next step 1230, it is assumed that the first step 1240 is based on a discounted price point 1222. As such, the value of the next step 1230 is treated as the base price for the time period of the first step 1241 (represented by the dashed line).
At this point, absolute discount (ΔP) and base price (BP) are used to calculate percent discount (ΔP/BP) for each store product time period. Percent discounts that are less than some value (e.g. 1%) are treated as being no discount and corrected to ΔP/BP=0. The above determined base price variable and percent discount variable are then output (Step 1213).
This base price is subjected to further analysis for accuracy using crossstore checking (Step 1215). This can be accomplished by analyzing the base price data for each product within a given store. A curve is generated for each product. This curve defines the price distribution for each product. The 80^{th }percentile for base price is then calculated for the analyzed product (i.e., the base price point below which 80% of the analyzed product (over the modeled time interval) is priced). This is referred to as the “in store 80^{th }percentile” for that product. A calculation is then made of the average 80^{th }percentile for price of the analyzed product across all stores (the crossstore 80^{th }percentile). Each store's prices are then merged with each other store to calculate the average 80^{th }percentile for base price over all stores.
The stores are then analyzed product by product. If the base price for a store is greater than two (2) standard deviations from the crossstore average 80^{th }percentile for base price and if the instore 80^{th }store percentile is more than 50% different from the crossstore 80^{th }percentile, this store is flagged as an outlier for the analyzed product.
Store  Product  In Store 80^{th }%  CrossStore 80^{th }%  Flagged 
Y  A  1.99  1.99  No 
Y  B  2.09  1.99  No 
Y  C  0.29  1.99  Yes 
Y  D  1.89  1.99  No 
The outlier store's base price is adjusted for the analyzed product such that it lies only two (2) standard deviations away from the average crossstore 80^{th }percentile for base price over all stores. This is illustrated in
Thus, the forgoing process illustrates an embodiment for determining an imputed base price variable.
b. Imputed Relative Price Variable
Reference is now made to the flowchart 1300 of
Equivalent  Actual  Equivalent  Equivalent  
Product  Size  Factor  Price  Units  Units  Price 
A  8  16  1.00  500  250  2.00 
B  16  16  2.00  300  300  2.00 
C  32  16  3.00  100  200  1.50 
Using this data, relative price may be calculated. As disclosed earlier, an equivalizing factor is defined. For this example, let the equivalizing factor be 16. Using the equivalizing factor, an equivalent price can be calculated (Step 1301).
Thus for
the results of these calculations are shown in the “Equivalent Price” column of the table above.
Next equivalent units sold (“units”) can be calculated (Step 1303).
Thus for
In a similar vein, equivalent base price and equivalent base units are calculated (Step 1305) using the imputed values for base price (for example, as determined in Steps 12011207) and for base units (also referred to as base volume which is determined as disclosed below).
For each Store, each demand group, and each date, the total equivalent units is determined (Step 1307). For example, using the dataset above (assuming that the data is from the same store), a total of 750 (i.e., 250+300+200) equivalent units were sold.
Defining A, B, and C as products in a demand group, the equivalent values for the demand group are depicted below:
Product  Equivalent Units  Equivalent Price 
A  250  $2.00 
B  300  $2.00 
C  200  $1.50 
A weighted calculation of relative equivalent price is then made (Step 1309). For example, such relative price value is determined as follows:
Equivalent price is divided by a weighted denominator.
The weighted denominator is calculated by multiplying equivalent units for each product times the equivalent units sold. For each product, only the values of other products are used in the calculation. This means excluding the product being analyzed. For example, if products A, B, and C are being analyzed in turn, when product A is analyzed the value for A is excluded from the denominator. Using the above data, the relative price of A is determined as follows:
To insure that all members of a demand group are counted at least at some minimal level, if equivalent units=0, a value of “1” is added to all units. In an example where equivalent units were A=0; B=5; C=1, the units would be revalued as A=1; B=6; C=12, and the calculations as disclosed above would be conducted. Also, where the number of products in a demand group is equal to one, the weighted average equivalent price for the single product is equal to the equivalent price for that product. If a value for equivalent price is missing, the equivalent price for the previous record is used for equivalent price.
Also, a weighted average equivalent base price is calculated using the method disclosed hereinabove. The only difference being that instead of using the actual equivalent price, the calculated base price values per equivalent are used (Step 1311). Using the previously disclosed techniques, a moving average is generated for relative actual equivalent price and relative equivalent base price (Step 1313).
This moving average is generally calculated by first defining a time period window framing each analyzed date (e.g., four weeks, two weeks prior, two weeks after). This framing time period is specified as an input. Second, for each date in the time period window, a weighted average of actual equivalent price and a weighted average of equivalent base price are calculated. For time period windows where there are insufficient days preceding the analyzed date (e.g., if the time window requires two week's worth of data but only one week is available), imputed values are provided for base price or actual price. Such imputed values are just the average value for base price or actual price, respectively. Third, once the time period window is defined, calculations are made defining average relative actual equivalent price and average relative equivalent base price over the time period window, thereby defining a moving average for both relative actual equivalent price and relative equivalent base price. This, is repeatedly done with the window being moved incrementally through the dataset thereby obtaining a moving average.
Thus a variety of imputed relative price variables can be generated (e.g., relative equivalent price, relative equivalent base price. etc.).
c. Imputed Base Volume Variable
A flowchart 1400 shown in
Using the nonpromoted data subset, an average value for “units” and a STD is calculated (i.e., an average value for product unit sales volume for each product during the nonpromoted dates is calculated) (Step 1405). The average units are rounded up to the nearest integer value, this value shall be referred to as the “nonpromoted average units”.
An initial value for base units (“initial base units”) is now determined (1407). This value is determined for all dates in the dataset, not just the nonpromoted dates. For those records having a percent price discount that is less than the predetermined discount level (e.g., 2%) the actual units sold are treated as “initial base units”. However, where such records (those the 2% or less discount) also have an actual value for units sold which is greater than 1.5 STD from the nonpromoted average unit value (as calculated above), then the actual value for units sold is not used. Instead, it is replaced with the nonpromoted average unit value in calculating “initial base units”. For the other records (those having a percent price discount that is equal to or greater than the predetermined discount level (e.g., 2%)), the previously calculated nonpromoted average unit value is used for “initial base units”.
This principle can be more readily understood with reference to
A calculated base volume value is now determined (Step 1409). This is accomplished by defining a time window. One preferred window is four (4) weeks, but the time window may be larger or smaller. For each store and product, the average value of “initial base units” is calculated for each time window. This value is referred to as “average base units”. This value is calculated for a series of time windows to generate a moving average of “average base units”. This moving average of the average base units over the modeled time interval is defined as the “base volume variable”.
d. Supplementary Error Detection and Correction
Based on previously determined discount information, supplementary error detection and correction may be used to correct price outliers. A flowchart 1500 illustrated in
The concepts are similar to that illustrated in
If the number of actual units sold (shown as 1560) lies within two (2) STD of the calculated base volume, then it is assumed that the actual price 1550 is actually an erroneous record and the actual value 1560 is replaced with the calculated base price.
However, if the number of actual units sold is greater than two (2) STD from the calculated base volume, it is assumed that the volume number is correct and the price is reset to reflect a discount of 60% and the price is recalculated based on the 60% discount. In short, the discount is capped at the chosen value (here 60%). Once the data is corrected, it can be output (step 1507).
e. Determining Imputed Variables which Correct for the Effect of Consumer Stockpiling
With reference to
“Lag” variables which define the number of product units sold (“units”) in the time leading up to the analyzed date are defined (Step 1607).
For example:
Lag1(units)=number of product units sold in one (1) time period (e.g., a day or week) before the analyzed date;
Lag2(units)=number of product units sold in two (2) time periods (e.g., a day or week) before the analyzed date;
Then the total number of product units sold is calculated for each defined time bucket (Step 1609). For example:
Bucket1=sum of units sold during the previous m days;
Bucket2=sum of units sold from between the previous m+1 days to 2m days;
Bucket3=sum of units sold from between the previous 2m+1 days to 3m days;
Bucket(τ)=sum of units from between the previous (τ−1)m+1 days to (τ)m days.
Correction can be made at the “front end” of the modeled time interval.
For example, the data can be viewed as follows:
Week  
1  Week 2  Week 3  Week 4  Week 5  Week 6  Week 7  
Bucket  —  Week 1  Week 2  Week 3  Week 4  Week 5  Week 6  
1  
Bucket  —  —  Week 1  Week 2  Week 3  Week 4  Week 5  
2  
Bucket  —  —  —  Week 1  Week 2  Week 3  Week 4  
3  
Bucket  —  —  —  —  Week 1  Week 2  Week 3  
4  
If working near the front end of a dataset, units from previous weeks cannot always be defined and in their place an averaged value for bucket sum can be used (Step 1611). For example, referring to Bucket 1, there is no Bucket 1 data for Week 1. As a result, the Bucket 1 data for weeks 27 are averaged and that value is put into Week 1 of Bucket 1. Similarly, with reference to Bucket 2, Week 1 and Week 2 are missing a value for Bucket 2, Weeks 13 are missing a value for Bucket 3, and Weeks 14 are missing a value for Bucket 4. The average values are generated for the missing values from weeks. For example, for Bucket 2, an average value for Weeks 37 is generated. This average value is used to fill out the missing dates of Bucket 2 (Weeks 12). Similarly, for Bucket 3, the average value for Weeks 47 are averaged and used to fill out the missing dates (Weeks 13). The same principle applies to Bucket 4. These Buckets define variables which are used to model the impact of promotional activity in previous time periods. The Buckets are used as variables in models which can be used to determine if there is a relationship between sales volume between a previous time as compared with a current time. The idea is to detect and integrate the effects of consumer stockpiling on into a predictive sales model.
f. Day of the Week Analysis
With reference to
Once categorized by day of the week the product units (sold) are summed for a specified dimension or set of dimensions. Dimension as used herein means a specified input variable including, but not limited to, Product, Brand, Demand Group, Store, Region, Store Format, and other input variable which may yield useful information (Step 1703). For example, if Region is the specified dimension (e.g., all the stores in Los Angeles, Calif.), all of the unit volume for selected products in the Los Angeles stores is summed for each Day of Week (i.e., 1, 2, 3, 4, 5, 6, and 7).
For each Day of Week and each dimension specified, the average units (sold) are determined (Step 1705). For each date, a “relative daily volume” variable is also determined (Step 1707). For example, relative daily volume for a given Store is provided by (total Store daily units)/(average Store daily units). Such calculation can be accomplished for any input variable.
One numeric example can be shown as follows. A store sells 700 units of X over a given modeled time interval. Average daily units=700/7=100. If sales for all of the Friday's of the modeled time interval amount to 150 units, it can be shown that, for that Store, Friday's relative daily volume is 1.5, i.e., more than average. This information may prove valuable to a client merchant and can comprise an input variable for other econometric models.
g. Imputed Seasonality Variable Generation
Another useful imputed variable is an imputed seasonality variable for determining seasonal variations in sales volume. One preferred approach for generating this variable is in accordance with the method described by Robert Blattberg and Scott Neslin in their book “Sales Promotion: Concepts, Methods, and Strategies”, at pages 237250 (Prentice Hall, N.J., 1990).
Referring to
The embodiment begins by categorizing the data into weekly data records. Chiefly, this comprises aggregating daily data into weekly groups (Step 1801). For missing sales records or records having zero volume values, insert average volume data (Step 1803).
A set of month variables is first defined (Step 1805). A series of models of base units are constructed using each defined month variable as the predictor.
The process of defining month variables is as follows:
Week  Month1  Month2  Month3  Month4 
1  1  12  12  12 
2  1  1  12  12 
3  1  1  1  12 
4  1  1  1  1 
5  2  1  1  1 
6  2  2  1  1 
7  2  2  2  1 
8  2  2  2  2 
. . .  . . .  . . .  . . .  . . . 
The values determined for base units are now processed. By taking the log of base units the effect of extreme variations in base units can be reduced (Step 1807). A linear regression is run on the log of base units for Month 1 (Step 1809). The regression models the log of base units as a function of Month 1 levels and Week number: (log(base units)=f(Month1, Week number)). The regression analysis is repeated using months Month2, Month3, and Month4 to determine, respectively log(base units)=f(Month2, Week number); log(base units)=f(Month3, Week number); and log(base units)=f(Month4, Week number).
h. Imputed Promotional Variable
Another useful variable is a variable which can predict promotional effects.
Referring back to
i. Imputed CrossElasticity Variable
Another useful variable is a crosselasticity variable.
An embodiment for generating crosselasticity variables in accordance with the principles of the present invention will be illustrated with reference to the following example. The subject category is an abbreviated soft drink category defined by demand groups for diet soft drinks (diet), regular cola soft drinks (reg), caffeine free soft drinks (cafffree), and root beer soft drinks (RB).
The initial dataset information is input into the system (Step 2001). For each demand group the total equivalent sales volume for each store is calculated for each time period (for purposes of this illustration the time period is a week) during the modeled time interval (Step 2003). For each week and each demand group, the average total equivalent sales volume for each store is calculated for each week over the modeled time interval (Step 2005). For each demand group the relative equivalent sales volume for each store is calculated for each week (Step 2007). This may be calculated for each store for each week in accordance with the formula below:
Relative Demand Group Equivalent Sales Volume=Total Equivalent Sales Volume For a Specific Week divided by Average Total Equivalent Sales Volume as Determined For All Weeks in The Modeled Time Interval.
The purpose of the crosselasticity variable is to quantify the effects of sales of one demand group upon the sales of another demand group. Therefore, when examining a first demand group, the sales of other demand groups within the same category are treated as variables which affect the sales of the first demand group. As such, the relative demand group equivalent sales volume for the other demand groups is quantified and treated as a variable in the calculation of sales volume of the first demand group, thereby generating crosselasticity variables (Step 2009). This can be illustrated more easily with reference to the partial dataset illustrated in Tables A and B. These tables reflect one week's data (week 1).
TABLE A  
DEMAND  RELATIVE DEMAND GROUP  
WEEK  PRODUCT  GROUP  EQUIVALENT VOLUME 
1  A  Diet 

1  B  Diet  ″ 
1  C  Diet  ″ 
1  D  Regular 

1  E  Regular  ″ 
1  F  Regular  ″ 
1  G  Cafffree 

1  H  Cafffree  ″ 
1  I  Cafffree  ″ 
1  J  RB 

1  K  RB  ″ 
1  L  RB  ″ 
TABLE B  
DEMAND  
PRODUCT  GROUP  CE_{Diet}  CE_{Regular}  CE_{Cafffree}  CE_{RB} 
A  Diet  —  X  X  X 
B  Diet  —  X  X  X 
C  Diet  —  X  X  X 
D  Regular  X  —  X  X 
E  Regular  X  —  X  X 
F  Regular  X  —  X  X 
G  Cafffree  X  X  —  X 
H  Cafffree  X  X  —  X 
I  Cafffree  X  X  —  X 
J  RB  X  X  X  — 
K  RB  X  X  X  — 
L  RB  X  X  X  — 
With reference to Table A it is shown that a calculation of Relative Demand Group Equivalent Volume for product A (a diet soda) is the total of all equivalent sales volume for the diet soda demand group for the time period (here week 1). This includes the sum of all equivalent sales volume for diet soda A, all equivalent sales volume for diet soda B, and all equivalent sales volume for diet soda C. This sum is divided by the average sum of equivalent sales volume for diet soda A, diet soda B, and diet soda C. This Relative Demand Group Equivalent Volume is a crosselasticity coefficient (CEdiet) for products other than diet soda (here, regular soda, caffeinefree soda, and root beer). The same type of calculation is performed with respect to regular soda (reg) and, for that matter, CaffeineFree (cafffree) and Root Beer (RB) as well. This yields four crosselasticity coefficients (CE_{diet}, CE_{reg}, CE_{cafffree}, and CE_{RB}). Table B illustrates the relationship between each product, demand group, and the four crosselasticity coefficients (CE_{diet}, CE_{reg}, CE_{cafffree}, and CE_{RB}). The crosselasticity coefficients are used generate crosselasticity variables for each product. In Table B the “” means the indicated crosselasticity coefficient is not applicable to the indicated product. An “x” means the indicated crosselasticity coefficient is applicable to the indicated product. For example, if product D (Regular soft drink) is examined, beginning with Table A, the equation for Relative Demand Group Equivalent Volume (for product A) is shown. This equation also yields the crosselasticity coefficient (CE_{reg}) for the demand group regular soda. Referring now to Table B, the row for product D is consulted. There are “x's” for the coefficients which apply to a determination of a crosselasticity variable for product D. Thus, crosselasticity for product D is a function of crosselasticity coefficients CE_{diet}, CE_{cafffree}, and CE_{RB}. Therefore, the crosselasticity variable for product D includes crosselasticity coefficients CE_{diet}, CE_{cafffree}, and CE_{RB}.
The calculated imputed variables and data are outputted from the imputed variable generator 304 to the coefficient estimator 308. Some of the imputed variables may also be provided to the financial model engine 108.
B. Coefficient Estimator
The coefficient estimator 308 uses the imputed variables and data to estimate coefficients, which may be used in an equation to predict demand. In a preferred embodiment of the invention, sales for a demand group (S) is calculated and a market share (F) for a particular product is calculated, so that demand (D) for a particular product is estimated by D=S·F. A demand group is defined as a collection of highly substitutable products. In the preferred embodiments, the imputed variables and equations for sales (S) of a demand group and market share (F) are as follows:
1. Modeling Framework
The econometric modeling engine relies on a mixedmodel framework, simultaneously utilizing information across all stores and products in a client category, where a category is defined as a collection of substitutable or complementary products. The mixed model methodology is also referred to as “Bayesian Shrinkage” Modeling, because by combining data from various stores and/or products, one can “shrink” individual parameter estimates towards the average estimate, dampening the extreme values that would result if traditional regression were used. A basic rationale for this approach is as follows.
In developing productlevel volume models for each store within a chain, one may be presented with a wide range of historical data in terms of modeling sufficiency. For some stores and/or products, the history will be quite rich, with many price changes and a variety of promotion patterns. For other stores and/or products, the history will be quite sparse, with very few price changes and little promotional activity. To maximize the stability of estimated model parameters, one might consider developing a single regression model across stores and/or products. This model might have stable parameter estimates; however, it would not fully leverage the store and/or product variability, and the resulting model would likely not predict well for a particular store and/or product. On the other hand, one might consider developing individual regression models for each store and/or product, to utilize the specific information contained in the data history. While these models might fit and predict well for the stores and/or products with substantial price variability, models would not be estimable for stores and/or products without a rich data history.
A mixedmodel framework addresses the need for both highly predictive models and the existence of an estimable model for each store and product. In a mixedeffect model, information (in the form of data history) is leveraged across all stores and products, and a single cohesive model is built. Stores and products with little or no information in their data history default to an “average” (fixedeffect) model. Likewise, stores and products with a wealth of information in their data history will end up with unique parameter estimates tailored to their response pattern, via estimation of nonzero store and/or productspecific adjustment factors (random effects) which are added to the fixedeffect portion of the model.
2. Terminology
The equivalent price of a product is defined as the price of a standardized unit of measure, which may be calculated based on the product description and the spread of sizes/counts that apply to that description. Each individual product price is divided by this standardized unit of measure to obtain the equivalent price.
A demand group is defined as a set of products that are substitutes or near substitutes for each other. A product can belong to only one demand group. A product category consists of one or more demand groups. For this example, attention is restricted to a single category consisting of multiple demand groups.
Both models:
Subscript i: Demand group (primary). A demand group is a collection of highly substitutable products.
Subscript j: Demand group (secondary). A secondary demand group is another demand group in the same category as the primary demand group, where a category is defined as a collection of substitutable or complementary products.
Subscript k: Product, where products are items with common UPC numbers.
Subscript t: Time period, which may be days, weeks, or hours.
Subscript B: Baseline, which is a state of product if there was no promotion.
Subscript n: Number of time periods away from current time period.
ε: Error term for regression equations, with appropriate subscripts per context.
3. Stage 1 (Sales) Model
a. Sales Model Multipliers (Data Values, or Covariates) and Dependent Variables
S_{i,j}: The equivalent sales of demand group i in period t in store s in dollars. Equivalent sales may be defined as sales of equivalent units of products being compared.
S_{Bi,t}: The equivalent baseline sales of demand group i in store s in period t.
R_{i,t}: The equivalent revenue of demand group i in period t.
R_{Bi,t}: The equivalent baseline revenue of demand group i in period t, which would be baseline sales times the baseline price.
P_{i,t}: The equivalent price of demand group i at store s in time period t, calculated as total equivalent revenue of demand group i divided by total equivalent sales in the period (S_{i,t}/R_{i,t}), and where the equivalent price is the price of an equivalent unit, such as a price per ounce.
M_{i,t}: The promotion level for demand group i in store s in period t.
X_{i,t}: The seasonality index for demand group i in store s in period t.
TS_{t}: The total dollar sales for the entire store in period t, computed using historical data.
T
b. Sales Model Factors (Parameters to be Estimated)
γ_{i}: The price elasticity factor for demand group i measured with respect to the deviations of the weighted average price of the demand group from the past weighted average price of the group. It measures the sensitivity of sales of equivalized units of the demand group with respect to the group price.
ν_{i}: The promotion factor for demand group i. This factor measures the sensitivity of the equivalent sales of the demand group to the general promotion level of the group.
ψ_{i}: The seasonality factor for demand group i. This factor measures the sensitivity of the equivalent sales of the demand group to seasonality.
κ_{i}: The seasonalityprice interaction factor that measures the interaction of weighted average price deviations and seasonality for demand group i. The seasonality and the group price may interact with each other in a nonlinear way. This factor measures a degree of nonlinearity.
δ_{i,n}: The time lag factor for demand group i and delay of n weeks. The time lag factor measures the level of forward buying or stockpiling activity by customers. Note that this is the only factor that is estimated at the demand group level rather than the storedemand group level.
φ_{i,j}: The cross elasticity factor for demand group i and demand group j. This factor measures how sales of a demand group are affected by the sales of other demand groups in the same category.
η_{i,t}: The competitive price factor for demand group i measured with respect to the difference between the weighted average price of the demand group within the store and outside competitors. This factor measures the effect of competitive activity on the sales of products in a given demand group.
π_{i}: The traffic factor for demand group i. Sales may be affected by the overall traffic through a store. This factor quantifies the relationship.
θ_{i}: The dayofweek (DOW) effect for demand group i. Each day of a week could have a different sales pattern. This factor quantifies the relationship.
K_{i}: The constant (intercept) associated with demand group i.
c. The Sales Model is:
In the above model (Equation 1), the dependent variable, demand group equivalent sales, is indexed (divided) by baseline demand group equivalent sales to provide a type of normalization. This normalizing allows easier comparison within a demand group and between demand groups. If a reasonable approximation of baseline demand group sales cannot be imputed, the dependent variable may alternatively be indexed by demand group equivalent sales averaged over a specified number of time periods prior to the current one (
In the time lag term, τ represents the number of “time buckets” preceding a time period that will be included in the model, and m represents the size of the “time bucket,” in number of time periods.
Inclusion of several covariates (dayofweek, store traffic, withinmarket competitive pricing) is contingent upon the time dimension and scope of available client data. Therefore, if data is reported on a weekly basis, so that there is no data according to day of the week, the day of the week parameters will not be included.
d. Sales Model Unbiasing Factor
In regression calculations, returning to the original scale after a logarithmic transformation creates a bias. To correct for this bias, the Baskersville's method is used which consists of adding an unbiasing factor to the equation. This factor is the mean square error ({circumflex over (σ)}^{2}) of the sales model divided by 2.
The equation for predicting demand group sales is thus:
4. Stage 2 (Share) Model
a. Share Model Multipliers (Data Values, or Covariates) and Dependent Variables
F_{i,k,t}: The fraction of demand group i equivalent sales comprised by product k in time period t (market share of product k).
P_{Bi,k,t}: The equivalent base price of product k in demand group i in time period t.
P_{R Bi,k,t}: The relative equivalent base price of product k in demand group i for time period
M_{p,i,k,t}: The level of promotion type p (kind of promotion) for product k in demand group i in time period t. There can be up to n_{p }promotion factors estimated in the model.
b. Share Model Factors (Parameters to be Estimated)
β_{i,k}: The relative base price elasticity factor for product k in demand group i.
σ_{p,i,k}: The promotion factor p for product k in demand group i. There can be up to n_{p }promotion factors estimated in the model.
χ_{i,k,n}: The time lag factor for product k in demand group i and delay of n weeks.
Λ_{i,k}: The constant (intercept) associated with product k in demand group I.
The model for predicting product share (market share) is:
This model calculates demand for a product divided by demand for the demand group of the product.
The product intercept Λ_{i,k }is not individually estimated in the model. Instead, each product is characterized according to product attributes, such as brand, size group (small/medium/large), form, flavor, etc. A storespecific estimate of the effect corresponding to each attribute level effects is obtained, and each product intercept is then constructed by summing over the applicable attribute level estimates.
Thus,
where
ξ_{a} _{ b }is the effect of attribute a, level b, and
I_{k,a} _{ k }is a is an indicator variable for product
The attribute values may be used to predict sales of new products with various attribute combinations.
5. Linearization of Multinomial Logistic Equation
The multinomial logistic function that defines the market share equations for each product is nonlinear but there exist standard techniques to transform it to a linear function instead, which may make modeling easier. An example of a transformation to a linear function is as follows:
In this section the store index is ignored.
For a particular time period t:
where α_{j }is the intercept for demand group i, β are the covariates, and P is the number of covariates in the share model
To model share in a linear framework, we simply center all covariates and model log(F_{i,k})−log({tilde over (F)}_{i}) where
C. Combined (Product Sales) Model
The model for predicting sales for product k in demand group i in time period t is thus given by:
{circumflex over (D)}_{i,k,t}={circumflex over (F)}_{i,k,t}Ŝ_{i,t }
III. Financial Model Engine
The financial model engine 108 receives data 132 from the stores 124 and may receive imputed variables (such as baseline sales and baseline prices) and data from the econometric engine 104 to calculate fixed and variable costs for the sale of each item. To facilitate understanding,
The financial model engine 108 should be flexible enough to provide a cost model for these different procedures. These different costs may have variable cost components where the cost of an item is a function of the amount of sales of the item and fixed cost components where the cost of an item is not a function of the amount of sales of the item. The financial model engine 108 uses these fixed and variable costs to determine C_{s,i,k,t}, where C_{s,i,k,t}, is a cost for a particular product (k) given a store (s), demand group (i), and a day (t). The financial model engine 108 uses industry data to provide standard estimates. For example, instead of measuring how long it takes to stock a box of an item, an industry data may be used to estimate this time. The standard estimates helps to reduce the amount of data that must be collected. In a preferred embodiment of the invention, the stores may only need to supply labor costs of the stores and distribution centers, cost of capital, size of an item, and number of items in a case to allow a cost modeling. Likewise, the preferred embodiment may infer the amount of shelfspace an item utilizes from the cubic feet of the item, the volume of sales of the item, and how often the store is replenished with the item. By using these estimations, costs may be more easily calculated on a store level, instead of being averaged over all of the stores. This is because if large amounts of data are measured by hand, measuring such data for each store would be difficult. The tailoring of costs per store allows the maximization of profits for each store. In an example of the preferred embodiment of the invention, the financial model engine 108 comprises an activitybased costing module that uses the parameters and calculations as follows:
A. Definitions
The Activitybased costing module computes variable and fixed costs for products at specific store locations. The ABC modeling component introduces the Distribution Center entity to track costing activities at retailers' distribution centers and regional stores.
Some numbers related to properties of the product. For each product a CasePack number is provided, which is the number of products in a case. A CaseCube number provides the cubic feet of each case of the product.
Certain specifiers are used to determine costs related to a distribution center. A ReceivingMethod specifier allows the specification of receiving the items by pallet or by individual cases, which specify different stocking costs. A ShelfStockMethod specifier allows the specification of the type of method that may by used to stock a shelf. For example, with baby food with many individual bottles, if the case provides a tray, which allows all bottles in the case to be quickly slid onto the shelf, then the stocking costs would be less than for a case without a tray, where all bottles must be manually lifted out of the case and placed on the shelf. If an entire box or case of an item may be placed on a shelf, the shelf stocking costs would also be less. A CaseListCost specifier specifies the wholesale price for a case. The CaseAllowance specifier specifies any discount off of the wholesale price of a case. The AllowBackhaul specifier allows the specification of a deduction provided by a manufacturer when a store is near the manufacturer so that on a return trip to a distribution center the store hauls the goods from the manufacturer. The BackhaulPerCase is the amount of the back haul discount per case. The VendorDaysCredit specifier allows the specification of the number of days of credit a manufacturer may give after shipment. The InventoryInDays specifier allows the specification of the number of days that the product is held as inventory at the distribution center.
If a wholesaler or manufacturer ships the product directly to the store, then different cost factors may be required, while distribution center related costs may be eliminated. The DropShipMethod specifier allows the specification of whether the directly shipped product is directly shipped to the shelf, a store back room or to a retailers distribution center.
The AvgWklyCase specifier specifies the average weekly number of cases of product sold by a store. This would be the average number of units of product sold by the store divided by the number of units of product per case.
Various costs may be associated with the process of having a product shipped directly to a store. The DropShipMethod specifier is used to specify whether a manufacturer delivers a product to a store shelf, or to a store back room, or to a distribution center.
Various labor costs may be associated with different distribution centers and stores. The DCLaborRate is a specifier used to specify the average labor rate at each distribution center. The StoreStockLaborRate is a specifier used to specify the average labor rate of a stock clerk at a store. The StoreCheckoutLaborRate is a specifier used to specify the average labor rate of a cashier at a store. Each of these labor rate averages may be different between different stores and different distribution centers. Each average per store or per distribution center may be used in the database.
A ProductStorageType specifier allows the specification of dry warehouse storage for dry items, or refrigerated storage, or frozen storage, which each have different costs. A DCDryCostPFT specifier is used to specify the cost of storing dry items, per cubic foot at the distribution center. This cost may be provided by the distribution center, but an industry average may be used as a default value. A DCRefrigCostPFT specifier is used to specify the cost of storing refrigerated items, per cubic foot at the distribution center. This cost may be provided by the distribution center, but an industry average may be used as a default value. A DCFrozenCostPFT specifier is used to specify the cost of storing frozen items, per cubic foot at the distribution center. This cost may be provided by the distribution center, but an industry average may be used as a default value. A StoreDryCostPFT specifier is used to specify the cost of storing dry items, per cubic foot at the store. This cost may be provided by the store, but an industry average may be used as a default value. A StoreRefrigCostPFT specifier is used to specify the cost of storing refrigerated items, per cubic foot at the store. This cost may be provided by the store, but an industry average may be used as a default value. A StoreFrozenCostPFT specifier is used to specify the cost of storing frozen items, per cubic foot at the store. This cost may be provided by the store, but an industry average may be used as a default value.
A DCCubeSecs specifier specifies the number of seconds it takes to handle a cubic foot of product at the distribution center. This time may be provided by the distribution center or an industry average may be used as a default value. A DCCaseUnitizedSecs specifier specifies the average number of seconds required to handle a palletized load at the distribution center. This time may be provided by the distribution center or an industry average may be used as a default value. A DCCaseDeadPileSecs specifier specifies the average number of seconds required to handle a nonpalletized load (a dead pile) at the distribution center. This time may be provided by the distribution center or an industry average may be used as a default value.
A StoreCubeNTSecs specifier specifies a time in seconds to handle a nontray cubic foot of product at a store. This time may be provided by the store or an industry average may be used as a default value. A StoreCaseNTSecs specifier is used to specify a time in unitized seconds to handle a nontray case of a product at a store. This time may be provided by the store or an industry average may be used as a default value. A StorePkgNTSecs specifier is used to specify a time in unitized seconds to handle a nontray package of a product at a store. This time may be provided by the store or an industry average may be used as a default value. So in a case with twenty four packages, when a package is purchased one twenty fourth of a case is purchased, which may have a volume of a tenth of a cubic foot.
A StoreCubeTraySecs specifier specifies a time in seconds to handle a cubic foot of a tray product at a store. This time may be provided by the store or an industry average may be used as a default value. A StoreCaseTraySecs specifier is used to specify a time in unitized seconds to handle a case of a tray product at a store. This time may be provided by the store or an industry average may be used as a default value. A StorePkgTraySecs specifier is used to specify a time in unitized seconds to handle a package of a tray product at a store. This time may be provided by the store or an industry average may be used as a default value.
A StoreCheckoutPkgSecs specifier specifies the time it takes to check out a package of product. This time may be provided by the store or an industry average may be used as a default value. An AvgDeliveryDistance is an average distance for shipping from a distribution center to a store. A TruckCubic specifier specifies the number of cubic feet that may be loaded on a truck. A StoreLaborPct specifier specifies the percentage of store labor that is required for the different delivery methods. If a distribution center delivers an item to a store, then the StoreLaborPct is 100%, since such products must be fully processed. If a vendor delivers a product directly to a backroom of the store, then a StoreLaborPct of 75% may be used, since the store does not need to do as much work. If a vendor delivers a product directly to a store shelf, then a StoreLaborPct of 25% may be used, since the store may be allowed to do less work for such deliveries.
An AnnualOpptyCost specifier specifies the annual opportunity cost of the inventory in the distribution centers and stores. Annual opportunity cost may be calculated by determining the inventory in the distribution centers and the inventory in the stores to determine the dollar value of the inventory and then subtract the dollar value of inventory that is on credit and multiplying the difference by the cost of capital percentage. Generally, the inventory may be calculated from the raw data, so that the customer only needs to provide the cost of capital to determine annual opportunity cost. An InvoiceProcessing specifier specifies the cost of providing an invoice (placing an order for a product with a vendor) per case. A Transportation specifier specifies a cost of transporting an item per mile. A ShoppingBag specifier specifies the cost of a shopping bag.
A database would provide an identification code for each distribution center and would provide for the specification of the location of each distribution center. Similarly, each store would have an identification code and the database would also allow the specification of the location of each store. A DeliveryFrequency specifier allows the specification of the number of days between deliveries to a store.
B. Cost Calculation
To calculate C_{s,i,k,t }(the cost of a product (i) in a demand group (k) in a store (s) at a time (t) fixed and variable cost components are computed from the given data structures, as follows:
The bag cost calculation computes total shopping bag costs for a product in a distribution center's region.
1. Bag Costs
The bag cost calculation is:
BagCost=(DistributionCenter.ShoppingBag/0.6)*Product.CaseCube/Product.CasePack,
where one bag in this example is 0.6 cubic feet, so that bag cost is the cost of a bag divided by the volume of a bag, the quantity times the volume of a case of a product divided by the number of products in a case. The distribution center identifier used in the equation is so that if bags have different costs at different distribution centers, the different cost may be noted by the distribution center identifier.
2. Checkout Labor Costs
The checkout labor costs is computed as cost per package from the rate at which items are checked out and the checkout clerk's pay rate. The Checkout Labor Cost Calculation is:
CheckoutCostPerPkg=StoreCheckoutPkgSecs*StoreCheckoutLaborRate/3600,
so that the Checkout Labor cost is the product of the store check out rate in packages per second and the store labor rate in dollars per hour divided by 3600 seconds per hour.
3. Distribution Center Inventory Costs
The Distribution Center Inventory Costs computes the inventory cost per package at a given distribution center. If a product is not stored at a distribution center, then the inventory cost is zero. If a product can be backhauled, the inventory cost is reduced by the backhaul rate. The Distribution Center Inventory Cost Calculation is as follows:
DCInvCostPerPkg=ISNULL(ProductDistributionCenter.InventoryInDays,0)*(DistributionCenter.AnnualOpptyCost/365)*(Product.CaseListCost−Product.CaseAllowance−IIf(Product.AllowBackHaul=1,Product.BackhaulPerCase,0))/Product.CasePack.
The zero in the first factor allows the cost to be zero if direct shipping is used. Otherwise, the inventory costs are the inventory costs for a distribution center times the annual opportunity cost for the distribution center divided by 365 days times the cost per case minus the allowances such as a product case allowance and back haul, with the back haul discount being zero if there is no back haul, where the product is divided by the number of products per case or package.
4. Distribution Center Labor Costs
The Distribution Center's Labor Costs per package is computed depending upon the distribution center's receiving method. If a product is not shipped from the distribution center, the cost is zero. The calculation of labor costs at the distribution center calculates the time it takes to process a product at the distribution center and multiplies it by a labor rate. An example of an algorithm for calculating distribution center labor costs is as follows:
Distribution Center Labor Cost Calculation
If Product.ReceivingMethod=1 Then
DCLaborCostPerPkg=((DistributionCenter.DCCubeSecs*Product.CaseCube/Product.CasePack)+(DistributionCenter.DCCaseUnitizedSecs/Product.CasePack))*(DistributionCenter.DCLaborRate/3600)
Else
DCLaborCostPerPkg=((DistributionCenter.DCCubeSecs*Product.CaseCube/Product.CasePack)+(DistributionCenter.DCCaseDeadPileSecs/Product.CasePack))*(DistributionCenter.DCLaborRate/3600)
In the first branch of the algorithm for a unitized (palletized) load, first the number of seconds to process a cubic foot of product processed by the distribution center is multiplied by the number of cubic feet per case divided by the number of products per case to obtain the number of seconds to process a product at the distribution center. The number of seconds to process a unitized (palletized) case by the distribution center divided by the number of products per case yields the number of seconds for each product being processed by the distribution center. These two products are added to obtain a total number of second to process a product processed by the distribution center. This total is multiplied by the distribution center labor rate per hour divided by 3600 seconds per hour to obtain the distribution center labor cost for palletized product.
In the second branch of the algorithm for a dead pile (nonpalletized) load, first the number of seconds to process a cubic foot of product processed by the distribution center is multiplied by the number of cubic feet per case divided by the number of products per case to obtain the number of seconds to process a product at the distribution center. The number of seconds to process a dead pile (nonpalletized) case by the distribution center divided by the number of products per case yields the number of seconds for each product being processed by the distribution center. These two products are added to obtain a total number of seconds to process a product processed by the distribution center. This total is multiplied by the distribution center labor rate per hour divided by 3600 seconds per hour to obtain the distribution center labor cost for nonpalletized products.
5. Distribution Center Calculations
The Distribution Center figures are aggregated by Product Type.
a. Sum of Average Weekly Cases
The SumOfAvgWklyCases specifier specifies the total of all average weekly cases of all stores for a particular product for a particular distribution center. This specifier may be calculated as follows:
SumOfAvgWklyCases=SUM(ProductLocDC.AvgWklyCases)
b. Count of Product Location Distribution Centers
The CountProdLocDC specifier specifies the number of stores per distribution center that sell the product.
c. Average Distribution Center Inventoried Days
The AvgOfDCInvDays specifier specifies the average number of days of inventory of a particular product at a distribution center. The distribution center may provide this number or an industry standard may be used as a default value.
d. Average Product Case Cubes
The AvgOfCaseCube specifier specifies an average cubic foot volume of a case of a product. This average may use an industrial standard as a default if a volume of a case of a product is not provided.
e. Distribution Center Space
The DCSpace specifier is a calculation of the volume of space of a product in the inventory of a distribution center. The calculation for DCSpace is as follows:
DCSpace=(AvgOfDCInvDays/7)*SumOfAvgWklyCase*AvgOfCaseCube
The DCSpace may be the product of the Average Distribution Center Inventoried Days divided by seven (to convert days into weeks) times the Sum of the Average Weekly Cases of the product times the Average Product Case Cube to yield the volume.
f. Distribution Center Cube
The DC Cube specifier specifies the volume of the distribution center allocated to the product. The distribution center may be divided into slots. In this embodiment, only whole numbers of slot spaces may be allocated to a product. In this example, the slot space is 80 cubic feet. Therefore, the volume used by inventory must be rounded up to the next 80 cubic feet increment. In such an example, an algorithm for calculating the number of slots allocated to the product is as follows:
If DC_Space>0 Then
DC_Cube=Int((DC_Space/80)+0.999)*80
Else
DC_Cube=0
In the first branch, if the Distribution Center Space is greater than zero, then the Distribution Center Space is divided by 80 cubic feet and the result is added to 0.999. The integer value of the resulting sum is used, so that the noninteger part of the sum is truncated. The integer value is multiplied by 80 cubic feet to convert the number of slots that are allocated to the product to a volume.
In the second branch, if the Distribution Center Space is equal to zero, no slots are allocated to the product.
g. Distribution Center Space Cost Per Store
The DCSpaceCostPerStore specifier specifies distribution center space cost per store. In this embodiment, an algorithm to calculate DCSpaceCostPerStore, where dry goods are ProductStorageType 1, refrigerated goods are ProductStorageType 2 and frozen goods are ProductStorageType 3, is as follows:
If ProductStorageType=1 Then
DCSpaceCostPerStore=DC_Cube*DCDryCostPFT/(Product.CasePack*52*CountProdLocDC)
If ProductStorageType=2 Then
DCSpaceCostPerStore=DC_Cube*DCRefrigCostPFT/(Product.CasePack*52*CountProdLocDC)
If ProductStorageType=3 Then
DCSpaceCostPerStore=DC_Cube*DCFrozenCostPFT/(Product.CasePack*52*CountProdLocDC)
In the first branch of the algorithm for dry goods, first the Distribution Center Cube is multiplied by the DCDryCostPFT. This product is then divided by the product of the number of product per case times 52 weeks in a year times the number of stores per distribution center to obtain the Distribution Center Space Cost Per Store.
In the second branch of the algorithm for refrigerated goods, first the Distribution Center Cube is multiplied by the DCRefrigCostPFT. This product is then divided by the product of the number of product per case times 52 weeks in a year times the number of stores per distribution center to obtain the Distribution Center Space Cost Per Store.
In the third branch of the algorithm for refrigerated goods, first the Distribution Center Cube is multiplied by the DCFrozenCostPFT. This product is then divided by the product of the number of product per case times 52 weeks in a year times the number of stores per distribution center to obtain the Distribution Center Space Cost Per Store.
6. Invoice Processing
The Invoice Processing calculation computes the Invoice processing cost per Product. To calculate Invoice Processing (HQInvPro), the cost of Invoice Processing per case is divided by the number of products per case, as follows:
HQInvProc=InvoiceProcessing/Product.CasePack
7. Product Location Space
The Product Location (store) Space calculation computes a store's available space for a product.
a. Average Delivery Frequency Calculation
The AvgDeliveryFreq specifier specifies the average number of deliveries to a store per week averaged over several weeks. An equation to calculate AvgDeliveryFreq is:
AvgDeliveryFreq=AVG(Location.DeliveryFrequency)
b. Sum of Average Weekly Cases Calculation
The SumOfAvgWklyCases specifier is the sum of the AvgWklyCases over several weeks.
SumOfAvgWklyCases=SUM(.AvgWklyCases)
c. Average Case Cube Calculation
The AvgCaseCube specifier specifies the average volume of a case of a product.
d. Location (Store) Cube Calculation
The LocationCube specifier specifies the volume of the store (location) that the product takes up on the shelf and/or back room. An equation for calculating LocationCube may be as follows:
LocationCube=(0.5*Product.CaseCube)+(Product.CaseCube*AvgWklyCases/DeliveryFrequency)
A volume of half a case is the first term. The second term is the volume of a case times the average number of cases delivered weekly divided by the frequency of deliveries per week, which estimates the volume of inventory delivered to the store with each delivery. So this sum estimates that the store will need a volume of about the volume for each delivery plus half a case to be used for the product.
e. Location Space Cost Calculation
After determining the volume of a product in the store, the store space cost (LocationSpaceCost) may be calculated. This cost is dependent upon whether the product is a dry item ProductStorageType=1, a refrigerated item ProductStorageType=2, or a frozen item ProductStorageType=3.
If ProductStorageType=1 Then
LocationSpaceCost=LocationCube*StoreDryCostPFT/(Product.CasePack*52)
If ProductStorageType=2 Then
LocationSpaceCost=LocationCube*StoreRefrigCostPFT/(Product.CasePack*52)
If ProductStorageType=3 Then
LocationSpaceCost=LocationCube*StoreFrozenCostPFT/(Product.CasePack*52)
In the first branch of the algorithm for dry goods, first the Location (store) Cube is multiplied by the StoreDryCostPFT to obtain a store storage cost per week. This product is then divided by the product of the number of product per case times 52 weeks in a year to obtain the Location Space Cost, which is the store storage cost per week per item of product.
In the second branch of the algorithm for refrigerated goods, first the Location (store) Cube is multiplied by the StoreRefrigCostPFT to obtain a store storage cost per week. This product is then divided by the product of the number of product per case times 52 weeks in a year to obtain the Location Space Cost, which is the store storage cost per week per item of product.
In the third branch of the algorithm for refrigerated goods, first the Location (store) Cube is multiplied by the StoreFrozenCostPFT to obtain a store storage cost per week. This product is then divided by the product of the number of product per case times 52 weeks in a year to obtain the Location Space Cost, which is the store storage cost per week per item of product.
8. Transportation Cost Per Package
The Transportation Cost Per Package (TransCostPerPkg) computes the cost to truck products to stores. Transportation Cost Per Package may be calculated as follows:
TransCostPerPkg=Product.CaseCube/(TruckCubic*Product.CasePack)*Transportation*AvgDeliveryDistance
The volume of the case is divided by the product of the volume of the truck times the number of items in each case. The result is multiplied by the cost per mile of delivery times the average delivery distance.
9. Product Location (Store) Inventory Cost
The Product Location Inventory Cost calculation computes a store's inventory cost for a product. An equation for calculating Product Location Inventory Cost may be as follows:
ProductLocInvCost=(((LocationCube*0.75/Product.CaseCube)/SumOfAvgWklyCases)−(VendorDaysCredit/7))*((Product.CaseListCost−Product.CaseAllowance−IIf(Product.AllowBackhaul=1,Product.BackhaulPerCase,0))/Product.CasePack)*AnnualOpptyCost/52
The volume of inventory in a store LocationCube is multiplied by 0.75. The 0.75 is an estimated adjustment since items are always being sold so that the actual inventory is a bit less than the actual volume of product delivered. This product is divided by the number of items per case. This result is divided by the Sum of the Average Weekly Cases to obtain an average cost per item. Then the Vendors Days Credit is divided by seven to get Vendor weeks credit, which is subtracted from the average cost per item. This difference is multiplied by the product list cost minus any allowances or discounts and back haul discounts and divided by the number of products per case to obtain the cost per product. This product is multiplied by the annual opportunity cost divided by 52 weeks in a year.
10. Location Receiving Stock Cost
The Location Receiving Stock Cost computes the labor cost to stock shelves. The cost depends upon tray and nontray stocking rates. Products are marked as tray or nontray stock depending upon the product's shelf stocking method.
An algorithm for calculating Location Receiving Stock is as follows:
If Product.ShelfStockingMethod 2 Then
LocationRcvgStkCost=((DistributionCenter.StoreCubeNTSecs*Product.CaseCube/Product.CasePack)+(DistributionCenter.StoreCaseNTSecs/Product.CasePack)+DistributionCenter.StorePkgNTSecs)*DistributionCenter.StoreStockLaborRate/3600
Else
LocationRcvgStkCost=((DistributionCenter.StoreCubeTraySecs*Product.CaseCube/Product.CasePack)+(DistributionCenter.StoreCaseTraySecs/Product.CasePack)+DistributionCenter.StorePkgTraySecs)*DistributionCenter.StoreStockLaborRate/3600
In the first branch of the algorithm for a nontray stocked item, first the time in seconds to handle a cubic foot of a nontray product at a store is multiplied by the cubic feet of each case of the product divided by the number of products per case to obtain the number of products per second of the nontray product is stocked on the shelves. The time in unitized seconds to handle a nontray case of a product at a store divided by the number of products per case yields the time per product of handling a product in stocking the shelves. These two products are added to obtain a total number of seconds per product for stocking the nontray product on the shelves. This total is multiplied by the store stocking labor rate per hour divided by 3600 seconds per hour to obtain the cost of stocking the nontray product on the shelf.
In the second branch of the algorithm for a tray stocked item, first the time in seconds to handle a cubic foot of a tray product at a store is multiplied by the cubic feet of each case of the product divided by the number of products per case to obtain the number of products per second of the tray product is stocked on the shelves. The time in unitized seconds to handle a tray case of a product at a store divided by the number of products per case yields the time per product of handling a product in stocking the shelves. These two products are added to obtain a total number of seconds per product for stocking the tray product on the shelves. This total is multiplied by the stocking labor rate per hour divided by 3600 seconds per hour to obtain the cost of stocking the tray product on the shelf.
11. Product Location Variable Cost Calculation
The ProdLocVariableCost specifier specifies variable cost for each product given a store, demand group, and day. In other embodiments of the invention the time period may be a week instead of a day. The ProdLocVariableCost specifier may be computed from the following equation:
ProdLocVariableCost=BagCost+ProductLocInvCost+CheckoutCostPerPkg+LocationRcvgStkCost+TransCostPerPkg+DCInvCostPerPkg+DCLaborCostPerPkg+HQInvProc
12. Fixed Cost Calculation
ProdLocFixedCost is the Fixed cost for each product given a store, demand group, and day, may be computed from a combination of costs computed above.
The fixed cost of a product may be calculated by the equation:
ProdLocFixedCost=LocationSpaceCost+DCSpaceCostPerStore
13. C_{s,i,k,t }Calculation
Where C_{s,i,k,t}, is a cost for a particular product (k) given a store (s), demand group (i), and a day (t), C_{s,i,k,t}, may be calculated as:
C _{s,i,k,t}=ProdLocVariableCost+ProdLocFixedCost.
IV. Price Optimization Engine and Support Tool
A. Overview
In operation, the client (stores 124) may access the rule editor 412 of the support tool 116 and provides client defined rule parameters (step 228). If a client does not set a parameter for a particular rule, a default value is used. Some of the rule parameters set by the client may be constraints to the overall weighted price advance or decline, branding price rules, size pricing rules, unit pricing rules, line pricing rules, and cluster pricing rules. These rules will be discussed in more detail regarding the preferred embodiment of the invention. The client defined parameters for these rules are provided to the rule tool 404 of the optimization engine 112 from the rule editor 412 of the support tool 116. Within the rule tool 404, there may be other rules, which are not client defined, such as a group sales equation rule. The rule parameters are outputted from the rule tool 404 to the price calculator 408. The demand coefficients 128 and cost data 136 are also inputted into the price calculator 408. The client may also provide to the price calculator 408 through the support tool 116 a desired optimization scenario rules. Some examples of scenarios may be to optimize prices to provide the optimum profit, set one promotional price and the optimization of all remaining prices to optimize profit, or optimized prices to provide a specified volume of sales for a designated product and to optimize price. The price calculator 408 then calculates optimized prices. The price calculator 408 outputs the optimized prices to the output display 416 of the support tool 116, which allows the stores 124 to receive the optimized pricing (step 232). In preferred embodiments, the rules and optimizing algorithm are as follows:
B. Preferred Embodiment of Optimization Module
The optimization engine uses the group sales equation and the market share equation previously defined in the section on the econometric engine to predict group sales and product market share, respectively. These two are then combined to predict product sales at the store level. The three equations are produced here:
1. First Stage Prediction
The predictive model for the group sales is:
2. Second Stage Prediction
The predictive model for estimating the fraction of group sales due to a product is:
The predictive model for demand for a given product is then given by {circumflex over (D)}_{s,i,k,t}={circumflex over (F)}_{s,i,k,t}Ŝ_{s,i,k,t}.
3. The Optimization Model
Regression models and the predictive equations derived from the above equations are used to construct the optimization model.
The objective is to maximize profit:
and the constraints are:
a. Obey the Regression Equations Governing
Ŝ_{i,t}∀i,k
b. Obey the Market Share Equations
c. Constrain Price Changes to be within a Given Range of Current Prices.
PMIN_{s,i,k,t}≦P_{s,i,k,t}≦PMAX_{s,i,k,t }
To simplify notation, the time index may be removed from the equations. Next, all constant terms in each of the expressions may be grouped together. So the objective becomes:
Maximize:
The regression equations for Ŝ_{i,t}∀i,k become
where
Similarly, the regression equations for the market share fractions become:
where
With this, the optimization problem may be written as:
4. Problem P1
Maximize:
Subject to
PMIN_{s,i,k}≦P_{s,i,k}≦PMAX_{s,i,k} c.
Now, in addition to the constraints listed above, the preferred embodiment may model several other business rules via constraints. These include limits on group price advance or decline, brand pricing rules, size pricing rules, and unit pricing rules.
d. Group Price Advance or Decline.
Since demand groups are made up of like or substitutable products, managers often wish to constrain the overall weighted price advance or decline in price for them, where the weights are the market shares of the products that constitute the demand groups. The constraints are:
e. Brand Pricing Rules:
Products are described and categorized by attributes such as brand, size and flavor. Therefore, each product may be associated with a set of attributes and values. These attributes are useful to us for several reasons. The attributes may be used in the regression modeling in order to create a parsimonious set of regression coefficients. They may also be used in setting rules. For instance, category managers might wish to offer store brands at lower prices compared to the competing national brands. They might also wish to constrain some product brands to be less expensive than others, either for strategic considerations or to meet their contractual obligations. Specifically, a manager can create a set Brand_{s,i}≡{(p_{s,i,l} _{ 1 },p_{s,i,k} _{ 2 }):p_{s,i,k} _{ 1 }must cost less than p_{s,i,k} _{ 2 }}, which leads to the following constraints:
∀s,i and (p _{s,i,k} _{ 1 } ,p _{s,i,k} _{ 2 })εBrand_{s,i} p _{s,i,k} _{ 1 } ≦p _{s,i,k} _{ 2 }
f. Size Pricing Rules:
Managers might also wish to create rules that relate the price of one product versus another based on their sizes. For instance, they might wish for products belonging to the same brand and sharing other attributes, but with different sizes to be priced such that the larger sized product costs less per equivalent unit of measure than a smaller one. They cSize_{s,i}≡{(p_{s,i,l} _{ 1 },p_{s,i,k} _{ 2 }):p_{s,i,k} _{ 1 }must cost less than p_{s,i,k} _{ 2 }an create a set which leads to the following constraints:
∀s,i and (p _{s,i,k} _{ 1 } ,p _{s,i,k} _{ 2 })εSize_{s,i} p _{s,i,k} _{ 1 } ≦p _{s,i,k} _{ 2 }
g. Unit Pricing Rules:
Continuing in the same vein, managers might wish to ensure that two products that are identical in every respect but size should be priced such that the larger product costs more than the smaller one. This rule is closely related to the size pricing rule. To implement this rule, managers can create a set Unit_{s,i}≡{(p_{s,i,l} _{ 1 },p_{s,i,k} _{ 2 }):e_{s,i,k} _{ 1 }p_{s,i,k} _{ 1 }must cost less than e_{s,i,k} _{ 2 }p_{s,i,k} _{ 2 }}, where e_{s,i,k }is the multiplicative factor to convert equivalent units into whole product units. This leads to the following constraints:
∀s,i and (p _{s,i,k} _{ 1 } ,p _{s,i,k} _{ 2 })εUnit_{s,i} e _{s,i,k} _{ 1 } p _{s,i,k} _{ 1 } ≦e _{s,i,k} _{ 2 } p _{s,i,k} _{ 2 }
h. Line Pricing Rules:
Retail customers expect common prices for certain groups of products such as, for example, cough lozenges of the same brand but different flavors. These rules may be classified as line pricing rules and implement them as follows. Product groups called line price groups may be defined as L_{1}, l=1, . . . ∥L∥, where every product within a line group must have the same price in a store; i.e.,
∀L _{1} ,l=1, . . . ∥L∥, and ∀k _{1} ,k _{2} εL _{1} P _{s,i,k} _{ 1 } =P _{s,i,k} _{ 2 }
i. Cluster Pricing:
Retailers define geographic and other regions within which they maintain the same price for a given product. This may be translated to the notion of store clusters. A store cluster Cluster_{c }is a set of stores such that the price of every product is invariant within the cluster. In other words, every product has the same price in every store within the cluster although each product may have a different price. In order to implement these constraints the set of stores may be partitioned into store clusters Cluster_{1}, . . . , Cluster_{c} where C≡{Cluster_{1}, . . . , Cluster_{C}}, and define a new price variable P_{c,i,k }which represents the common price for product k in cluster Cluster_{c}. This variable may be used in place of the original price variable P_{s,i,k }whenever sεClusters. The entire problem P1 can be rewritten as P2:
5. Problem P2:
Maximize:
(Profit Maximizing Objective)
Subject to:
PMIN_{c,i,k}≦P_{c,i,k}≦PMAX_{c,i,k},∀i,k (Product price advance/decline) d.
∀c,i and (p _{c,i,k} _{ 1 } ,p _{c,i,k} _{ 2 })εBrand_{c,i} p _{c,i,k} _{ 1 } ≦p _{c,i,k} _{ 2 }(Brand Pricing) e.
∀c,i and (p _{c,i,k} _{ 1 } ,p _{c,i,k} _{ 2 })εSize_{c,i} p _{c,i,k} _{ 1 } ≦p _{c,i,k} _{ 2 }(Size Pricing) f.
∀c,i(p _{c,i,k} _{ 1 } ,p _{c,i,k} _{ 2 })εUnit_{c,i} ,sεCluster_{c} ,e _{s,i,k} _{ 1 } p _{c,i,k} _{ 1 } ≦e _{s,i,k} _{ 2 } p _{c,i,k} _{ 2 }(Unit Pricing) g.
∀L _{1} ,l=1, . . . ∥L∥,∀k _{1} ,k _{2} εL _{1} ,∀c,P _{c,i,k} _{ 1 } =P _{c,i,k} _{ 2 }(Line Pricing) h.
The optimization problem P2 has the following features: the objective and the group sales equations are nonlinear, while the rest of the constraints are linear. In the preferred embodiment, a heuristic approach is used to generate a good feasible starting point for the problem, if one exists. The heuristic works by repeatedly solving a related problem involving all the linear constraints 48, a set of linear constraints that approximate the group price advance and decline constraints (3), and an objective that involves the marginal contribution to profit of each product.
6. Heuristic to Generate the Feasible Solution:
The problem may be decomposed by store cluster and hence attention may be restricted to a single cluster Cluster_{c}. for this explanation.
In the preferred embodiment, the heuristic focuses on finding good initial feasible solutions to the problem. First, a related problem is defined.
The market shares of all products are calculated based on a set of initial prices. Then these initial market shares are used as fixed weights to develop linear constraints that limit the advance and decline of demand group prices at each store. Specifically, let
the market share of product k at store s given price vector
Note that the weights G_{s,i,k}(
7. Problem P3(
Subject to:
PMIN_{c,i,k}≦p_{c,i,k}≦PMAX_{c,i,k},∀i,k (Product price advance/decline) 2.
∀i and (p _{c,i,k} _{ 1 } ,p _{c,i,k} _{ 2 })εBrand_{c,i} p _{c,i,k} _{ 1 } ≦p _{c,i,k} _{ 2 }(Brand Pricing) 3.
∀i and (p _{c,i,k} _{ 1 } ,p _{c,i,k} _{ 2 })εSize_{c,i} p _{c,i,k} _{ 1 } ≦p _{c,i,k} _{ 2 }(Size Pricing) 4.
∀i,(p _{c,i,k} _{ 1 } ,p _{c,i,k} _{ 2 })εUnit_{c,i} ,sεCluster _{c} ,e _{s,i,k} _{ 1 } ≦e _{s,i,k} _{ 2 } p _{c,i,k} _{ 2 }(unit Pricing) 5.
∀L _{1} ,l=1, . . . ∥L∥,∀k _{1} ,k _{2} εL _{1} ,P _{c,i,k} _{ 1 } =P _{c,i,k} _{ 2 }(Line Pricing) 6.
Problem P3 has a nonlinear objective but only linear constraints. Note that the Problem P3 does not include group sales. The objective maximizes the sum of marginal profit for all products over all the stores in the cluster. This approach may be easily extended to include a multiplier that approximates the group sales of all groups over all stores.
To facilitate understanding,
and
This ensures that
In practice, convergence is declared in step 2 when the change in prices is below an acceptable tolerance threshold.
This method turns out to have several advantages. First, it finds an initial feasible solution by solving iteratively a nonlinear problem with linear constraints. Problem P3 is considerably easier to solve than Problem P1 and Problem P2. Second, the initial feasible solution turns out to be a good starting solution to Problem P2. It converges quite quickly from this starting point.
Therefore these initial feasible solutions are provided as a starting point for solving Problem P2.
An example of this would have two products A and B. Each would be currently priced in the store at $1. In this example the rules only provide one constraint, which in this example is the Group price advance/decline constraint so that
Therefore, for step 604, the current incumbent prices are set equal to the current price in the store ($1). Using the demand and market share models, the market share for each product A and B is calculated for the given incumbent prices of $1. For example, the market share of A may be calculated to be 50% (0.5) and the market share of B may be calculated to be 50% (0.5). Therefore, the constraint
may be approximated by PMIN_{s,i}≦F_{A}P_{A}+F_{B}P_{B}≦PMAX_{s,i}, which yields the constraint equation PMIN_{s,i}≦(0.5)P_{A}+(0.5)P_{B}≦PMAX_{s,i }(Step 606). The present incumbent prices of $1, the linear constraint equation, and the nonlinear objective of maximizing profit are solved with a nonlinear optimization problem software package to find optimized prices when the constraint equation is PMIN_{s,i}≦(0.5)P_{A}+(0.5)P_{B}≦PMAX_{s,i }(Step 608). An example of such nonlinear optimization problem software packages are MINOS™, which was invented by Bruce Murtagh and Michael Saunders and is marketed by Stanford Business Software, Inc, and CONOPT™, which was written by Arne Stolbjerg Drud and marketed by ARKI Consulting and Development A/S.
In this example the nonlinear optimization software package provides an optimized price for product A to be $0.70 and an optimized price for product B to be $1.20. The current incumbent prices are compared to the optimized price. Since for product A $1≠$0.70 or for product B $1≠$1.20 (Step 620), then the incumbent prices are set equal to the optimized price (Step 624). From the new incumbent prices new market shares are calculated, using the model for predicting product share. For example, the new calculated market share may be 65% (0.65) for Product A and 35% (0.35) for Product B. The new market shares are placed into the constraint equation to provide a new constraint equation (Step 606), such as PMIN_{s,i}≦(0.65)P_{A}+(0.35)P_{B}≦PMAX_{s,i}. The present incumbent prices of $0.70 and $1.20, the linear constraint equation, and the nonlinear objective of maximizing profit are solved with a nonlinear optimization problem software package to find optimized prices when the constraint equation is PMIN_{s,i}≦(0.65)P_{A}+(0.35)P_{B}≦PMAX_{s,i }(Step 608).
In this example the nonlinear optimization software package provides an optimized price for Product A to be $0.70 and an optimized price for Product B to be $1.20. Since the optimized prices are now equal to (or are within a threshold delta) the incumbent prices (Step 620), the group price bonds are reset so that the optimized prices for Product A ($0.70) and Product B ($1.20) are feasible for Problem P2 (Step 628). The optimized prices are placed in the demand equation to generate group sales (Step 632). The optimized prices and group sales are then used as a feasible starting solution to Problem P3 (Step 636). This means that the optimized prices, or Product A ($0.70) and Product B ($1.20), the generated group sales, the Profit Optimization Object, and the constraints of Problem P2 are provided to a nonlinear optimization problem software, such as MINOS or CONOPT. The use of the optimized prices allows for a quick convergence and an initial feasible solution for the solving of Problem P2. The initial feasible solution is important because in a nonlinear equation there may be several local optimal locations and there are several areas where a solution is not feasible or there is no convergence or convergence is very slow. The initial feasible solution provides a feasible local optimal location where there is a faster convergence.
In general, optimizations with a linear objective and linear constraints are the easiest to solve. Optimizations with nonlinear objectives and linear constraints may be harder to solve. Optimizations with nonlinear objectives with nonlinear constraints may be the hardest to solve. Therefore the heuristic uses an iterative approximation to convert an optimization with a nonlinear objective and nonlinear constraints to an optimization with a nonlinear objective and linear constraints to provide an initial solution. Once the initial solution is solved, the initial solution is used as a starting point to optimize a nonlinear objective with nonlinear constraints. Such optimizations may provide a gradient around the initial solution to determine a direction in which to go. Prices are used in that direction to obtain a next point. A gradient around the new prices is then used to determine a new direction. This may be iteratively done until a boundary is reached or until the prices do not change. Other optimization heuristics and other approaches may be used in different embodiments of the invention.
The use of calculating demand group demand as a function of price and then using market share to calculate a product's demand from the demand group demand in an optimization scheme is a novel aspect of the invention.
CPU 922 is also coupled to a variety of input/output devices, such as display 904, keyboard 910, mouse 912 and speakers 930. In general, an input/output device may be any of: video displays, track balls, mice, keyboards, microphones, touchsensitive displays, transducer card readers, magnetic or paper tape readers, tablets, styluses, voice or handwriting recognizers, biometrics readers, or other computers. CPU 922 optionally may be coupled to another computer or telecommunications network using network interface 940. With such a network interface, it is contemplated that the CPU might receive information from the network, or might output information to the network in the course of performing the abovedescribed method steps. Furthermore, method embodiments of the present invention may execute solely upon CPU 922 or may execute over a network such as the Internet in conjunction with a remote CPU that shares a portion of the processing.
In addition, embodiments of the present invention further relate to computer storage products with a computerreadable medium that have computer code thereon for performing various computerimplemented operations. The media and computer code may be those specially designed and constructed for the purposes of the present invention, or they may be of the kind well known and available to those having skill in the computer software arts. Examples of computerreadable media include, but are not limited to: magnetic media such as hard disks, floppy disks, and magnetic tape; optical media such as CDROMs and holographic devices; magnetooptical media such as floptical disks; and hardware devices that are specially configured to store and execute program code, such as applicationspecific integrated circuits (ASICs), programmable logic devices (PLDs) and ROM and RAM devices. Examples of computer code include machine code, such as produced by a compiler, and files containing higher level code that are executed by a computer using an interpreter.
Examples of Price Optimization
In an example of the invention, using a single category for a chain of stores, the category had 430 products, 6 demand groups, and 433 stores. Table 1 shows the number of products in each demand group. Within each demand group products compete against each other, such as an 8 oz. Coke and a 24 oz. Coke. Within the category products may compete or may be complementary. An example of a category may be drinks. In this example, 12 months of scanner data was collected from the stores to provide the econometric demand model coefficients. In this example costs were not directly accounted for, but instead profit was measured by the difference between selling price and the unit cost.
TABLE 1  
Demand Group  
A  B  C  D  E  F  
Number of Products  5  44  18  17  74  21  
The econometric demand model coefficients were calculated by the use of the SAS™ program provided by SAS Institute Inc., of Cary N.C. An error analysis was performed on the demand model to determine a confidence interval for one standard deviation (68%). Table 2 shows the error for a one standard deviation confidence interval. Pct. Error is measured as 100*(forecasted value−actual value)/(forecasted value), and Abs. Pct. Error is calculated as the absolute value of Pct. Error.
TABLE 2  
Error  Error  Error  
Weighted  Weighted  Weighted  
Average  Standard  Average  
Level  N  Type  Mean  Deviation  Median 
Store Product  22821  Pct.Error (Volume)  −13.22%  46.20%  −1.97% 
Store Product  22821  Abs.Pct.Error (Volume)  47.54%  32.61%  38.48% 
Store Product  22821  Pct.Error (Profit)  −13.26%  46.59%  −2.01% 
Store Product  22821  Abs.Pct.Error (Profit)  47.62%  32.97%  38.52% 
Store Demand  2361  Pct.Error(Volume)  −0.45%  14.11%  −1.06% 
Group  
Store Demand  2361  Abs.Pct.Error (Volume)  12.87%  9.34%  12.12% 
Group  
Store Demand  2361  Pct.Error (Profit)  −0.57%  17.44%  0.52% 
Group  
Store Demand  2361  Abs.Pct.Error (Profit)  14.34%  12.28%  12.99% 
Group  
Store Category  394  Pct.Error (Volume)  0.15%  7.27%  0.32% 
Store Category  394  Abs.Pct. Error (Volume)  6.27%  4.77%  5.86% 
Store Category  394  Pct.Error (Profit)  0.87%  7.59%  1.52% 
Store Category  394  Abs.Pct.Error (Profit)  6.80%  5.02%  6.41% 
Table 2 shows that at the Store Category level, the error weighted standard deviation in volume of sales between predicted sales and actual sales was 7.27%. This means that there is a 68% confidence (one standard deviation) that the actual sales would fall within 7.27% of predicted sales or in other words the probability that the actual price would be more than one standard deviation away from the predicted price would be 32%. Examining the Store Demand Group level the error weighted standard deviation in volume increases to 14.11%. At the product level, the error weighted standard deviation in volume increases to 46.20%. Generally, the stores in this example carry extra inventory as safety stock, maybe as long as 4 weeks. It has been estimated that the 46.20% error would be covered by the general safety stock policies in most stores.
In this example group prices were allowed to vary between 85% and 105% of their current prices and the product prices were allowed to vary between 70% and 115% of their current levels. Therefore in this example there were only two constraints, which were:
PMIN_{s,i}
(Group Pricing—approximate) with PMIN being 85% of current prices and PMAX being 105% of current prices.
Profit was then optimized to provide new optimized prices, using the two constraints. This optimization is of profit is measured at the category level, since aggregated profit is of more interest than profit by demand group. This is because a store may be willing to have lower profits for a single demand group if the profit of an entire category is raised.
In the specification, examples of product are not intended to limit products covered by the claims. Products may for example include food, hardware, software, real estate, financial devices, intellectual property, raw material, and services. The products may be sold wholesale or retail, in a brick and mortar store or over the Internet, or through other sales methods.
V. Rule Relaxation
As mentioned above, various rules may be used as constraints for optimization. The rules described above for price optimization are:
PMIN_{c,i,k}≦P_{c,i,k}≦PMAX_{c,i,k},∀i,k (Product price advance/decline) 2.
∀c,i and (p _{c,i,k} _{ 1 } ,p _{c,i,k} _{ 2 })εBrand_{c,i} p _{c,i,k} _{ 1 } ≦p _{c,i,k} _{ 2 }(Brand Pricing) 3.
∀c,i and (p _{c,i,k} _{ 1 } ,p _{c,i,k} _{ 2 })εSize_{c,i} p _{c,i,k} _{ 1 } ≦p _{c,i,k} _{ 2 }(Size Pricing) 4.
∀c,i,(p _{c,i,k} _{ 1 } ,p _{c,i,k} _{ 2 })εUnit_{c,i} ,sεCluster _{c} ,e _{s,i,k} _{ 1 } p _{c,i,k} _{ 1 } ≦e _{s,i,k} _{ 2 } p _{c,i,k} _{ 2 }(Unit Pricing) 5.
∀L _{1} ,l=1, . . . ∥L∥,∀k _{1} ,k _{2} εL _{1} ,∀c,P _{c,i,k} _{ 1 } =P _{c,i,k} _{ 2 }(Line Pricing) 6.
Other rules may also be used. An example of such additional rules may be, but is not limited to, a gross margin rule, a store volume rule, or a competition rule.
A gross margin rule provides a constraint on the change of a mark up. A gross margin rule may provide a constraint on the change of the mark up of each individual item in a group or on the average mark up of the group or both. An example of a gross margin rule may be that the percentage change of the average gross margin of a group is not to exceed 5%.
A volume rule provides a constraint on the change or the volume of sales. The volume rule may provide a constraint on the change of volume of sales for each individual store or on the change of total volume of sales for a plurality of stores for an individual product or for a plurality of products. An example of a volume rule may be that for each store, the change of demand for a product is not to exceed 7%.
A competition rule provides a constraint on the difference between a competing store's prices. A competition rule may allow constraints for one or more competitors. An example of a multiple competition rule is that the price of each item must not exceed the price for the same item sold by competitor store X by more than 10% of the price and must not exceed the price of the same item sold by competitor store Y by more than 9% of the price. To implement these constraints a store may use monitors to check prices of items in competitor store X and competitor store Y.
During an optimization, it may be found that a rule is infeasible. A rule is determined to be infeasible, if either it cannot be satisfied when all other rules are satisfied or if an optimization cannot be performed without violating the rule. One method of dealing with an infeasible rule is to drop the rule found to be infeasible.
To provide an example of rule relaxation, six categories of rules are provided. According to the method illustrated in the flow chart of
1. Group price advance or decline rules. The user sets a maximum weighted group price advance or decline to 10%.
2. Size pricing rules. The user goes with the default that larger items cost less per equivalent unit than smaller identical items.
3. Brand pricing rules. For soft drinks, the user designates that the price of Brand A is never less than the price of Brand B. For juices the user designates that Brand C is always greater than Brand D.
4. Unit pricing rules. The user goes with the default that the overall price of larger items is greater than the overall price of smaller identical items.
5. Competition rules. The user designates that all prices must be at least 10% less than the prices of the same items sold by competitor X and are within 2% of the prices of the same items sold by competitor Y.
6. Line price rules. The user designates that different flavors of the same item are priced the same.
In this example, Brand A is sold by competitor Y for $8, while Brand B is sold by competitor Y for $10. In addition competitor Y sells one flavor of item C for $9 and another flavor of the same item C for $10.
A determination is made if a rule is infeasible (step 2108). It is found that one of the price branding rules is infeasible, since Brand A is supposed to always have a price greater than the price of Brand B and yet competitor Y sells Brand A for $8 and Brand B for $10, and since the competition rules designate that the same items must be within 2% of the prices of competitor Y. In addition, it is found that one of the line price rules is infeasible. This is because the line price rule constrains different flavors of the same item to be the same, but competitor Y is selling one flavor for $9 and another flavor for $10.
Since the branding rules have a priority of 3 and the line price rules have a priority of 6, it is found the line price rules are the lowest priority infeasible (LPI) rules (step 2112). A check is made to see if rules with a priority lower than 6 may be relaxed so that the line price rules become feasible (step 2116). Since the line price rules are the lowest priority rules, no rule with a lower priority may be relaxed so that the line price rule becomes feasible. As a result, the line price rule is relaxed allowing the different flavors to have different prices by about 10% (step 2124).
The process then returns to step 2108 to see if any rules are still infeasible. It is found that one of the price branding rules is infeasible, since Brand A is supposed to always have a price greater than the price of Brand B and yet competitor Y sells Brand A for $8 and Brand B for $10, and since the competition rules designate that the same items must be within 2% of the prices of competitor Y. Since the price branding rules are the only rules found to be infeasible, they are found to be the lowest priority infeasible (LPI) rules with a priority of 3 (step 2112). A check is made to see if rules with priorities lower than the LPI rules may be relaxed to allow the LPI to be feasible (step 2116). Preferably, the lower priority rules are checked first. So first the line price rules with a priority of 6 are checked to see if they may be relaxed to allow the branding price rules to be feasible. If not, then the next lowest priority rules, the competition rules are checked to see if they may be relaxed to allow the branding rules to be feasible. In this example, it is found by relaxing a competition rule with a priority of 5 the price branding rule becomes feasible. Therefore the competition rule is relaxed so that prices are within about 20% of the prices of competitor Y, which allows Brand A and Brand B to have the same price (step 2120).
The process again returns to step 2108 to see if any rules are still infeasible. It is found that all rules are now feasible, and the rule relaxation process is stopped (step 2128).
In the preferred implementations, the specific infeasible solution at which the optimization process stops will depend on the optimization algorithm and the initial starting point. This means that specific rules that are relaxed in order to get to feasibility depend on both the optimization algorithm and the starting values These examples are merely illustrative and are not meant to limit the scope of the invention. In addition, rule priority was assigned by rule category. In other embodiments the rule priority may be assigned to individual rules instead of to rule categories.
Other specific methods of rule relaxation may be used. In general, rule relaxation methods provide for the prioritization of rules and cause lower priority rules to be relaxed before relaxing higher priority rules in order to make all resulting rules feasible. Another specific embodiment of this general rule relaxation method may allow different combinations of lower priority rules to be relaxed. These lower priority rules may be relaxed at the same rate or at different rates until the higher priority rule becomes feasible. Different combinations of the relaxation of lower priority rules may be specified in different embodiments.
While this invention has been described in terms of several preferred embodiments, there are alterations, modifications, permutations, and substitute equivalents, which fall within the scope of this invention. It should also be noted that there are many alternative ways of implementing the methods and apparatuses of the present invention. It is therefore intended that the following appended claims be interpreted as including all such alterations, modifications, permutations, and substitute equivalents as fall within the true spirit and scope of the present invention.
Cited Patent  Filing date  Publication date  Applicant  Title 

US3017610  Mar 15, 1957  Jan 16, 1962  Curtiss Wright Corp  Electronic data file processor 
US4744026  Apr 11, 1986  May 10, 1988  American Telephone And Telegraph Company, At&T Bell Laboratories  Methods and apparatus for efficient resource allocation 
US4862357 *  Nov 18, 1988  Aug 29, 1989  Systemone Holdings, Inc.  Computer reservation system with means to rank travel itineraries chosen in terms of schedule/fare data 
US4907170  Sep 26, 1988  Mar 6, 1990  General Dynamics Corp., Pomona Div.  Inference machine using adaptive polynomial networks 
US4916443  Oct 27, 1988  Apr 10, 1990  Supra Products, Inc.  Method and apparatus for compiling data relating to operation of an electronic lock system 
US5063506  Oct 23, 1989  Nov 5, 1991  International Business Machines Corp.  Cost optimization system for supplying parts 
US5117354  Jun 5, 1990  May 26, 1992  Carnes Company, Inc.  Automated system for pricing and ordering custom manufactured parts 
US5189606  May 14, 1991  Feb 23, 1993  The United States Of America As Represented By The Secretary Of The Air Force  Totally integrated construction cost estimating, analysis, and reporting system 
US5212791 *  Jul 27, 1992  May 18, 1993  International Business Machines Corporation  Dynamic scheduling 
US5249120  Jan 14, 1991  Sep 28, 1993  The Charles Stark Draper Laboratory, Inc.  Automated manufacturing costing system and method 
US5299115  Feb 5, 1993  Mar 29, 1994  Mrs. Fields Software Group Inc.  Product demand system and method 
US5377095  Jul 7, 1992  Dec 27, 1994  Hitachi, Ltd.  Merchandise analysis system with sales data table and various functions for predicting the sale by item 
US5459656  Feb 26, 1993  Oct 17, 1995  Park City Group, Inc.  Business demand projection system and method 
US5521813  Jan 15, 1993  May 28, 1996  Strategic Weather Services  System and method for the advanced prediction of weather impact on managerial planning applications 
US5615109 *  May 24, 1995  Mar 25, 1997  Eder; Jeff  Method of and system for generating feasible, profit maximizing requisition sets 
US5694551  Apr 24, 1995  Dec 2, 1997  Moore Business Forms, Inc.  Computer integration network for channeling customer orders through a centralized computer to various suppliers 
US5712985  Oct 13, 1995  Jan 27, 1998  Lee; Michael D.  System and method for estimating business demand based on business influences 
US5732401  Mar 29, 1996  Mar 24, 1998  Intellitecs International Ltd.  Activity based cost tracking systems 
US5765143  Mar 10, 1995  Jun 9, 1998  Triad Systems Corporation  Method and system for inventory management 
US5774868  Dec 23, 1994  Jun 30, 1998  International Business And Machines Corporation  Automatic sales promotion selection system and method 
US5790643  Feb 28, 1995  Aug 4, 1998  British Telecommunications Public Limited Company  Pricing method for telecommunication system 
US5799286  Jun 7, 1995  Aug 25, 1998  Electronic Data Systems Corporation  Automated activitybased management system 
US5822736  Feb 28, 1995  Oct 13, 1998  United Hardware Distributing Company  Variable margin pricing system 
US5832456  Jan 18, 1996  Nov 3, 1998  Strategic Weather Services  System and method for weather adapted, business performance forecasting 
US5832458  Nov 19, 1997  Nov 3, 1998  Electronic Data Systems Corporation  System and method for electronically auditing pointofsale transactions 
US5873069  Oct 13, 1995  Feb 16, 1999  American Tv & Appliance Of Madison, Inc.  System and method for automatic updating and display of retail prices 
US5878400  Jun 17, 1996  Mar 2, 1999  Trilogy Development Group, Inc.  Method and apparatus for pricing products in multilevel product and organizational groups 
US5902351  Aug 24, 1995  May 11, 1999  The Penn State Research Foundation  Apparatus and method for tracking a vehicle 
US5918209  Jan 11, 1996  Jun 29, 1999  Talus Solutions, Inc.  Method and system for determining marginal values for use in a revenue management system 
US5933813  Apr 15, 1996  Aug 3, 1999  Eldat Communication Ltd.  Sales promotion data processor system and interactive changeable display particularly useful therein 
US5987425  Oct 10, 1997  Nov 16, 1999  United Hardware Distributing Company  Variable margin pricing system 
US6009407  Feb 27, 1998  Dec 28, 1999  International Business Machines Corporation  Integrated marketing and operations decisionsmaking under multibrand competition 
US6009415  Dec 16, 1991  Dec 28, 1999  The Harrison Company, Llc  Data processing technique for scoring bank customer relationships and awarding incentive rewards 
US6029139  Jan 28, 1998  Feb 22, 2000  Ncr Corporation  Method and apparatus for optimizing promotional sale of products based upon historical data 
US6032123  Apr 29, 1998  Feb 29, 2000  Jameson; Joel  Method and apparatus for allocating, costing, and pricing organizational resources 
US6032125  Oct 22, 1997  Feb 29, 2000  Fujitsu Limited  Demand forecasting method, demand forecasting system, and recording medium 
US6044357  May 5, 1998  Mar 28, 2000  International Business Machines Corporation  Modeling a multifunctional firm operating in a competitive market with multiple brands 
US6052686  Sep 18, 1997  Apr 18, 2000  At&T Corporation  Database processing using schemas 
US6078893  May 21, 1998  Jun 20, 2000  Khimetrics, Inc.  Method for stabilized tuning of demand models 
US6094641 *  May 21, 1998  Jul 25, 2000  Khimetrics, Inc.  Method for incorporating psychological effects into demand models 
US6125355  Dec 2, 1997  Sep 26, 2000  Financial Engines, Inc.  Pricing module for financial advisory system 
US6134534  Sep 4, 1997  Oct 17, 2000  Priceline.Com Incorporated  Conditional purchase offer management system for cruises 
US6173345  Nov 3, 1998  Jan 9, 2001  Intel Corporation  Method and apparatus for levelizing transfer delays for a channel of devices such as memory devices in a memory subsystem 
US6202070  Dec 31, 1997  Mar 13, 2001  Compaq Computer Corporation  Computer manufacturing system architecture with enhanced software distribution functions 
US6205431  Oct 29, 1998  Mar 20, 2001  Smart Software, Inc.  System and method for forecasting intermittent demand 
US6219649  Jul 6, 1999  Apr 17, 2001  Joel Jameson  Methods and apparatus for allocating resources in the presence of uncertainty 
US6308162 *  May 21, 1998  Oct 23, 2001  Khimetrics, Inc.  Method for controlled optimization of enterprise planning models 
US6321207  Apr 15, 1999  Nov 20, 2001  I2 Technologies Us, Inc.  System and method for optimizing the allocation of a resource 
US6341268  Dec 23, 1998  Jan 22, 2002  Walker Digital, Llc  System and method providing a restaurant menu dynamically generated based on revenue management information 
US6341269  Dec 30, 1999  Jan 22, 2002  Mercani Technologies, Inc.  System, method and article of manufacture to optimize inventory and merchandising shelf space utilization 
US6397193  Jan 22, 1998  May 28, 2002  Walker Digital, Llc  Method and apparatus for automatically vending a combination of products 
US6405175  Jul 27, 1999  Jun 11, 2002  David Way Ng  Shopping scouts web site for rewarding customer referrals on product and price information with rewards scaled by the number of shoppers using the information 
US6430539  May 6, 1999  Aug 6, 2002  Hnc Software  Predictive modeling of consumer financial behavior 
US6456986  Jul 15, 1999  Sep 24, 2002  American Management Systems, Incorporated  Decision network based event pricing system in a component based, object oriented convergent customer care and billing system 
US6536935 *  Jul 23, 1997  Mar 25, 2003  Atarum Institute  Computerized system for marketbased constraint optimization 
US6546387  Nov 15, 1999  Apr 8, 2003  Transcom Software Inc.  Computer network information management system and method using intelligent software agents 
US6553352  May 4, 2001  Apr 22, 2003  Demand Tec Inc.  Interface for merchandise price optimization 
US6567824  Jun 20, 2001  May 20, 2003  Grantley Patent Holdings, Ltd.  Integrated inventory management system 
US6609101  Mar 25, 2000  Aug 19, 2003  The Retail Pipeline Integration Group, Inc.  Method and system for determining timephased product sales forecasts and projected replenishment shipments for a retail stores supply chain 
US6684193  Oct 5, 1999  Jan 27, 2004  Rapt Technologies Corporation  Method and apparatus for multivariate allocation of resources 
US6697824  Aug 31, 1999  Feb 24, 2004  Accenture Llp  Relationship management in an Ecommerce application framework 
US6725208  Apr 12, 1999  Apr 20, 2004  Pavilion Technologies, Inc.  Bayesian neural networks for optimization and control 
US6731998  Mar 20, 2003  May 4, 2004  I2 Technologies Us, Inc.  Collaboratively solving an optimization problem using first and second optimization software each having at least partial information concerning the optimization problem 
US6735572  Mar 13, 2002  May 11, 2004  Mark Landesmann  Buyerdriven targeting of purchasing entities 
US6745184  May 31, 2001  Jun 1, 2004  Rosetta Marketing Strategies Group  Method and system for clustering optimization and applications 
US6826538 *  Jul 28, 1999  Nov 30, 2004  I2 Technologies Us, Inc.  Method for planning key component purchases to optimize revenue 
US6910017  Mar 5, 1999  Jun 21, 2005  Profitlogic, Inc.  Inventory and price decision support 
US6934931  Apr 5, 2001  Aug 23, 2005  Pavilion Technologies, Inc.  System and method for enterprise modeling, optimization and control 
US6965867  Dec 13, 1999  Nov 15, 2005  Joel Jameson  Methods and apparatus for allocating, costing, and pricing organizational resources 
US6988076  Sep 10, 2001  Jan 17, 2006  Khimetrics, Inc.  Strategic planning and optimization system 
US7058617  Sep 14, 2000  Jun 6, 2006  Pavilion Technologies, Inc.  Method and apparatus for training a system model with gain constraints 
US7062447  Dec 20, 2000  Jun 13, 2006  Demandtec, Inc.  Imputed variable generator 
US7072848  Nov 15, 2001  Jul 4, 2006  Manugistics, Inc.  Promotion pricing system and method 
US7092896  May 4, 2001  Aug 15, 2006  Demandtec, Inc.  Interface for merchandise promotion optimization 
US7092918  May 4, 2001  Aug 15, 2006  Demandtec, Inc.  Apparatus for merchandise price optimization 
US7092929  Jul 13, 2001  Aug 15, 2006  Bluefire Systems, Inc.  Method and apparatus for planning analysis 
US7130811  May 5, 2001  Oct 31, 2006  Demandtec, Inc.  Apparatus for merchandise promotion optimization 
US7155402  Nov 8, 2000  Dec 26, 2006  Bluefire Systems, Inc.  Method and apparatus for distribution of fashion and seasonal goods 
US7240019  May 10, 2002  Jul 3, 2007  Demandtec Inc.  Interface for merchandise price optimization 
US7249031 *  Nov 26, 2001  Jul 24, 2007  Demandtec Inc.  Selective merchandise price optimization 
US7249032  Nov 30, 2001  Jul 24, 2007  Demandtec Inc.  Selective merchandise price optimization mechanism 
US7249033  Nov 30, 2001  Jul 24, 2007  Demandtec Inc.  Apparatus and method for selective merchandise price optimization 
US7302410 *  Dec 22, 2000  Nov 27, 2007  Demandtec, Inc.  Econometric optimization engine 
US7349879 *  Dec 1, 2000  Mar 25, 2008  Alsberg Peter A  Methods and systems for market clearance 
US20010014868 *  Jul 22, 1998  Aug 16, 2001  Frederick Herz  System for the automatic determination of customized prices and promotions 
US20020023001  Jul 11, 2001  Feb 21, 2002  Mcfarlin James A.  Method and apparatus for online creation and sale of custom local store marketing materials 
US20020042739  Mar 13, 2001  Apr 11, 2002  Kannan Srinivasan  Method and system for creating and administering internet marketing promotions 
US20020042755 *  Oct 4, 2001  Apr 11, 2002  I2 Technologies, Us, Inc.  Collaborative fulfillment in a distributed supply chain environment 
US20020107819  Sep 10, 2001  Aug 8, 2002  Ouimet Kenneth J.  Strategic planning and optimization system 
US20020116348  May 18, 2001  Aug 22, 2002  Phillips Robert L.  Dynamic pricing system 
US20020123930  Nov 15, 2001  Sep 5, 2002  Manugistics Atlanta Inc.  Promotion pricing system and method 
US20020169657  Oct 29, 2001  Nov 14, 2002  Manugistics, Inc.  Supply chain demand forecasting and planning 
US20020184059 *  Mar 25, 1999  Dec 5, 2002  Joseph Robert Offutt  Methods and apparatus for determining nonobvious savings in the purchase of goods and services 
US20020198794  Jun 21, 2001  Dec 26, 2002  Williams Edward P.  Inventory management system for determining suggested part stocking levels for a vehicle dealer 
US20030110072  May 4, 2001  Jun 12, 2003  Demand Tec. Inc.  Interface for merchandise promotion optimization 
US20030177103  Oct 23, 2002  Sep 18, 2003  I2 Technologies Us, Inc.  Calculating price elasticity 
US20040111358  Aug 13, 2003  Jun 10, 2004  Jeffrey Lange  Enhanced parimutuel wagering 
US20040210541  May 1, 2003  Oct 21, 2004  Jeremy Epstien  User interface for a rules engine and methods therefor 
US20040243432  Jul 12, 2004  Dec 2, 2004  Revenue Management Systems, Inc.  Method for assigning retail units to economic markets 
US20050108070  Oct 26, 2004  May 19, 2005  Thomas Kelly  Method for assigning retail units to economic markets 
US20060161504  Mar 20, 2006  Jul 20, 2006  I2 Technologies Us, Inc.  Generating an optimized price schedule for a product 
US20060224534  Apr 3, 2006  Oct 5, 2006  Hartman Eric J  Method and apparatus for training a system model with gain constraints using a nonlinear programming optimizer 
WO1997046950A1  May 29, 1997  Dec 11, 1997  Multex Systems, Inc.  Information delivery system and method 
WO1998053415A1  May 21, 1998  Nov 26, 1998  Khimetrics, Inc.  A method for incorporating psychological effects into demand models 
WO2000070519A2  May 12, 2000  Nov 23, 2000  From2.Com, Inc.  Network accessible quotation and shipping system 
WO2000070556A2  May 19, 2000  Nov 23, 2000  Whitehead Institute For Biomedical Research  A method and relational database management system for storing, comparing, and displaying results produced by analyses of gene array data 
WO2002001456A1 *  Jun 20, 2001  Jan 3, 2002  Dealtime.Com Ltd.  Ecommerce real time demand and pricing system and method 
Reference  

1  "Essentus and Spotlight Solutions Partnership Delivers Precise Markdown Decisions", Business Wire, Apr. 17, 2000, 3 pages.  
2  "Goodyear Implements Trilogy's MultiChannel Pricing Solution as Its EnterpriseWide EPricer Platform." Business Wire, p. 2286, Jun. 27, 2000.  
3  "Gymboree Enhances Price Management," Retail Systems Alert, vol. 13, No. 6, Jun. 2000.  
4  "IMRglobal Signs New Product Implementation Agreement with Retek to Improve Retail Pricing and Markdown Process." Business Wire, p. 1590, Jun. 7, 2000.  
5  "KhiMetrics and Retek Form Marketing Alliance for Comprehensive Retail Pricing Solution." PR Newswire, Feb. 19, 2001.  
6  "KhiMetrics Helps Retailers Increase Margins With Two New Tools for Their Retail Revenue Management Application Suite." PR Newswire, Mar. 1, 2001.  
7  "Manugistics Agrees to Acquire Talus Solutions." PR Newswire, Sep. 21, 2000.  
8  "Merriam Webster's Collegiate Dictionary", 10th edition, p. 585, MerriamWebster Incorporated, 1999.  
9  "New Tools for Modeling Elasticity, Optimizing Prices and Collecting Live Pricing from the Web to Debut at Retail Systems 2000 in Chicago." Business Wire, p. 1484, Apr. 18, 2000.  
10  "Pacificorp IRP: Renewables Costs Must Drop 65% to be Competitive with Gas".  
11  "Pacificorp IRP: Renewables Costs Must Drop 65% to be Competitive with Gas," dated Dec. 8, 1995.  
12  "PCT International Search Report", Application No. PCT/US02/14977, mailed May 5, 2003.  
13  "PCT International Search Report", Application No. PCT/US02/36710, mailed Jul. 21, 2003.  
14  "PCT International Search Report", Application No. PCT/US03/30488, mailed Jan. 28, 2004.  
15  "Report of Novelty Search" by Patentec, dated Feb. 9, 2001.  
16  "Report of Novelty Search" by Patentec, dated Jul. 25, 2001.  
17  "Supplementary European Search Report", Application No. 02719197.22221 dated Feb. 2, 2007.  
18  A.F.M. Smith, "A General Bayesian Linear Model," University of Oxford, Apr. 1972.  
19  Abraham, Magid M. And Leonard M. Lodish, "Promoter: An Automated Promotion Evaluation System," Marketing Science, vol. 6, No. 2, 1987 (p. 101123).  
20  Alan L. Montgomery and Peter R. Rossi, "Estimating Price Elasticities with TheoryBased Priors," Journal of Marketing Research vol. XXXVI, Nov. 1999 (pp. 413423).  
21  Alan L. Montgomery, "Creating MicroMarketing Pricing Strategies Using Supermarket Scanner Data," Marketing Science, vol. 16, No. 4, 1997 (pp. 315337).  
22  Alan Mercer, "Nonlinear Price Effects", Journal of the Market Research Society, dated Jul. 1, 1996, p. 227.  
23  Andrew B. Gelman et al., "Bayesian Data Analysis", pp. 439455, Chapman & Hall/CRC, First Edition 1995, Reprinted 2000.  
24  Arnold Zellner, "An Efficient Method of Estimating Seemingly Unrelated Regressions and Tests for Aggregation Bias," University of Wisconsin, Jun. 1962.  
25  Arnold Zellner, "On Assessing Prior Distribution sand Bayesian Regression Analysis With GPrior Distributions," Elsevier Science Publishers, 1986 (pp. 233243).  
26  Barth, Brad, "ShopKo Holds the Price Line." Daily News Record, p. 10, Oct. 4, 2000.  
27  Barth, Brad, "Shopko Tests Automated Markdowns", WWD Oct. 4, 2000, pp. 13.  
28  BernersLee, T., "Hypertext Markup Language 2.0 Working Paper," Nov. 1995 (pp. 13).  
29  Binkley, James K.; Connor, John M., "Grocery Market Pricing and the New Competitive Environment." Journal of Retailing, v74, n2, Summer 1998.  
30  Blattberg and Deighton, "Manage Marketing by the Customer Equity," Harvard Business Review, Jul.Aug. 1996, pp. 136144.  
31  Boatwright, Peter et al., "AccountLevel Modeling for Trade Promotion: An Application of a Constrained Parameter Hierarchical Model," Journal of the American Statistical Association, vol. 94, No. 448, Dec. 1999 (pp. 10631073).  
32  Bruce G.S. Hardie et al., "Attributebased Market Share Models: Methodological Development and Managerial Applications", University of Pennsylvania, Working Paper 98009, pp. 148, 1998.  
33  Bruce G.S. Hardie et al., "Attributebased Market Share Models: Methodological Development and Managerial Applications", University of Pennsylvania, Working Paper 98009, pp. 148.  
34  Bucklin & Gupta, "Brand Choice, Purchase Incidence, and Segmentation: An Integrated Modeling Approach," Journal of Marketing Research, May 1992, pp. 201215, vol. XXIX.  
35  Busch: "Cost Modeling as a Technical Management Tool", ResearchTechnology Management, Nov./Dec. 1994, vol. 37, No. 6, pp. 5056.  
36  Buzzell, Quelch, & Salmon; "The Costly Bargain of Trade Promotion;" Harvard Business Review, Reprint 90201, Mar.Apr. 1990, pp. 19.  
37  ByungDo Kim et al., "Modeling the Distribution of Price Sensitivity and Implications for Optimal Retail Pricing", Journal of Business & Economic Statistics, Jul. 1995, vol. 13, No. 3.  
38  Cerf, Vinton G. And Robert E. Kahn, "A Protocol for Packet Network Interconnection," IEEE Transactions on Communications COM22, May 1974, (p. 637648).  
39  Christen, Gupta, Porter, Staelin & Wittink; "Using MarketLevel Data to Understand the Effectiveness of Promotional Activities;" Dec. 22, 1995.  
40  Cook, Martie, "Optimizing Space and Sales with Markdown Software", Office.com, May 31, 2000, p. 1.  
41  Curry, Divakar, Mathur & Whiteman; "Bvar as a Category Management Tool: An Illustration and Comparison with Alternative Techniques;" Journal of Forecasting, vol. 14, Iss. No. 3 (1995), pp. 181199.  
42  D.V. Lindley and A.F.M. Smith, "Bayes Estimates for the Linear Model," University College, Dec. 1971.  
43  *  Dembeck et al. , "Using a Spreadsheet to Solve a Mutiinational Marketing Problem", Jan 1997, Management Accounting (USA), v78, n7, p33(7). 
44  Diebold, Francis X., "The Past, Present, and Future of Macroeconomic Forecasting" Economic Perspectives (IJEP), vol. 12, No. 2, pp. 175192, Spring 1998.  
45  Dyer, Robert F. et al., "Case Studies in Marketing Decisions Using Expert Choice," Decision Support Software, 1988, pp. 27, 73108.  
46  Eileen Bridges et al., "A HighTech Product Market Share Model With Customer Expectations" Marketing Science, vol. 14, No. 1, Winter 1995, pp. 6181.  
47  Flanagan, David, "Javascript: The Definitive Guide, 3rd Edition," published by O'Reilly in Jun. 1998 (ISBN 1565923928) section 14.8.  
48  Francis J. Mulhern et al., "The Relationship between Retail Price Promotions and Regular Price Purchases", Journal of Marketing, vol. 59, Oct. 1995, pp. 8390.  
49  Fusun Gonul, "Modeling Multiple Sources of Heterogeneity in Multinomial Logit Models: Methodological and Managerial Issues", Marketing Science, vol. 12, No. 3, Summer 1993, pp. 213229.  
50  Fusun Gonul, "Modeling Multiple Sources of Heterogeneity in Multinomial Logit Models: Methodological and Managerial Issues", Marketing Science, vol. 12, No. 3, Summer 19932, pp. 213229.  
51  George C. Tiao and Arnold Zellner, "On the Bayesian Estimation of Multivariate Regression," University of Wisconsin, Jan. 1964.  
52  Hernandez, Mauricio A., and Salvatore J. Stolfo, "Realworld Data is Dirty: Data Cleansing and the Merge/Purge Problem", Data Mining and Knowledge Discovery, vol. 2, Issue 1, Jan. 1998.  
53  Hillier, Frederick S., et al. "Introduction to Operations Research", McGrawHill, Inc., 1995, Sixth Edition, pp. 114.  
54  Jagmohan S. Raju, "The Effect of Price Promotions on Variability in Product Category Sales", Marketing Science, vol. 11, No. 3, Summer 1992, pp. 207220.  
55  John Deighton et al., "The Effects of Advertising on Brand Switching and Repeat Purchasing", Journal of Marketing Research, vol. XXXI, Feb. 1994, pp. 2843.  
56  John Philip Jones, "The Double Jeopardy of Sales Promotions," Harvard Business Review, Sep.Oct. 1999, pp. 145152.  
57  Kadiyali et al., "Manufacturerretailer Channel Interactions and Implications for Channel Power: An Investigation of Pricing in Local Market", Marketing Science, Spring 2000, V. 19, Issue 2.  
58  Lee G. Cooper et al., "Standardizing Variables in Multiplicative Choice Models", Journal of Consumer Research, vol. 10, Jun. 1983, pp. 96108.  
59  Little, John D. C., "Brandaid: A MarketingMix Model, Part 1: Structure," Operations Research, vol. 23, No. 4, Jul.Aug. 1975 (p. 628655).  
60  Magid M. Abraham et al., "An Implemented System for Improving Promotion Productivity Using Store Scanner Data", Marketing Science, vol. 12, No. 3, Summer 1993.  
61  Melcer, Rachel, "Local Tech Firm Creates Retail Markdown Tool", Business Courier online, Mar. 24, 2000, pp. 14.  
62  Montgomery: "The Impact of MicroMarketing on Pricing Strategies", 1994 The University of Chicago vol. 55/12A of Dissertation of Abstracts International, p. 3922 (Abstract Only).  
63  Mulhern et al. "Measuring Market Response to Price Changes: A Classification Approach". J BUSN RES 1995:33.197205.  
64  Peter M. Guadagni et al., "A Logit Model of Brand Choice Calibrated on Scanner Data", Marketing Science, vol. 2, No. 3, Summer 1983, pp. 203238.  
65  Peter S. Fader et al., "Modeling Consumer Choice among SKUs", Journal of marketing Research, vol. XXXII (Nov. 1996), pp. 442452.  
66  Peter S. Fader et al., "Modeling Consumer Choice among SKUs", Journal of Marketing Research, vol. XXXIII (Nov. 1996), pp. 442452.  
67  Richard A. Briesch, "Does it Matter How price Promotions Are Operationalized?", Marketing Letters 8:2 (1997), pp. 167181.  
68  Richard R. Batsell, et al., "A New Class of Market Share Models", Marketing Science, vol. 4, No. 3, Summer 1985, pp. 177198.  
69  Robert C. Blattberg and Edward I. George, "Shrinkage Estimation of Price and Promotional Elasticities: Seemingly Unrelated Equations," Journal of the American Statistical Association, vol. 86, No. 414, Jun. 1991 (pp. 304315).  
70  Robert C. Blattberg et al., "How Promotions Work", Marketing Science, vol. 14, No. 3, Part 2 of 2, 1995, pp. G122G132.  
71  Robert J. Dolan, "How Do You Know When the Price is Right?", Harvard Business Review, Sep.Oct. 1995, p. 411.  
72  Robert J. Dolan, "How Do You Know When the Price is Right?", Harvard Business Review, Sep.Oct. 1995, pp. 511.  
73  Robert M. Schindler et al., "Increased Consumer Sales Response through Use of 99Ending Prices", Journal of Retailing, Jun. 1, 1996, p. 187.  
74  Rockney G. Walters, "Assessing the Impact of Retail Price Promotions on Product Substitution, Complementary Purchase, and Interstore Sales Displacement", Journal of Marketing, vol. 55, Apr. 1991, pp. 1728.  
75  Ross Link, "Are Aggregate Scanner Data Models Biased?," Journal of Advertising Research, Sep./Oct. 1995, pp. RC8RC12, ARF.  
76  Rossi Delorgio, & Kantor; "Making Sense of Scanner Data;" Harvard Business Review, Reprint F00205.  
77  Rossi, Delorgio, & Kantor; "Making Sense of Scanner Data;" Harvard Business Review, Reprint F00205, 2000.  
78  Russell & Kamakura, "Understanding Brand Completion Using Micro and Macro Scanner Data," Journal of Marketing Research, vol. XXXI (May 1994), pp. 289303.  
79  Scherage, Dan, "You Do the Math," Chain Store Age, v76, n7, Jul. 2000.  
80  Smith et al., "A Discrete Optimization Model for Seasonal Merchandise Planning." Journal of Retailing, vol. 74, No. 2, p. 193(29), Summer 1998.  
81  Smith, Mathur, & Kohn; Bayesian Semiparametric Regression: An Exposition and Application to Print Advertising;: Jan. 3, 1997; Australian Graduate School of Management, University of New South Wales, Sydnet 2025, Australia.  
82  Stephen J. Hoch at al., "Determinants of StoreLevel Price Elasticity", Journal of Marketing Research, vol. XXXII (Feb. 1995), pp. 1729.  
83  Stephen J. Hoch et al., "Store Brands and Category Management", The Wharton School, University of Pennsylvania, Mar. 1998, pp. 138.  
84  Sunil Gupta, "Reflections on 'Impact of Sales Promotions on When, What, and How Much to Buy'", Journal of Marketing Research, vol. XXX, Nov. 1993, pp. 522524.  
85  Technology Strategy Incorporated, www.grossprofit.com, Mar. 2, 2000, pp. 120.  
86  Tellis, Gerard J., and Fred S. Zufryden, "Tackling the Retailer Decision Maze: Which Brands to Discount, How Much, When and Why," Marketing Science, vol. 1, No. 34, 1995 (pp. 271299).  
87  William R. Dillon et al., "A Segmentlevel Model of Category Volume and Brand Choice", Marketing Science, vol. 15, No. 1, 1996, pp. 3859.  
88  Yoeman, John Cornelius Jr., "The Optimal Offering Price for Underwritten Securities", 1993, vol. 55/01A of Dissertation Abstracts International, p. 4743 (Abstract Only). 
Citing Patent  Filing date  Publication date  Applicant  Title 

US7813948 *  Oct 12, 2010  Sas Institute Inc.  Computerimplemented systems and methods for reducing cost flow models  
US7895067 *  Feb 22, 2011  Oracle International Corp.  Systems and methods for optimizing total merchandise profitability  
US7945496 *  May 17, 2011  Pricemetrix, Inc.  Reference price framework  
US7996331 *  Aug 9, 2011  Sas Institute Inc.  Computerimplemented systems and methods for performing pricing analysis  
US8000996  Nov 8, 2007  Aug 16, 2011  Sas Institute Inc.  System and method for markdown optimization 
US8108249 *  Jan 31, 2012  KimberlyClark Worldwide, Inc.  Business planner  
US8160917  Apr 17, 2012  Sas Institute Inc.  Computerimplemented promotion optimization methods and systems  
US8185414 *  May 22, 2012  Medikredit Integrated Healthcare Solutions (Proprietary) Limited  System and method of processing a health insurance claim  
US8200518  Jun 12, 2012  Sas Institute Inc.  Computerimplemented systems and methods for partial contribution computation in ABC/M models  
US8234155 *  Nov 30, 2007  Jul 31, 2012  Sap Ag  Method and system incorporating business rules and price optimization for financial system 
US8271318  Sep 18, 2012  Sas Institute Inc.  Systems and methods for markdown optimization when inventory pooling level is above pricing level  
US8311878  Nov 13, 2012  Oracle International Corporation  Optimized pricing solver with prioritized constraints  
US8364510  May 24, 2011  Jan 29, 2013  Vuelogic, Llc  Revenue optimization for customers or customer subsets 
US8494887 *  Dec 9, 2002  Jul 23, 2013  Jda Software Group, Inc.  Generating an optimized pricing plan 
US8515835  Aug 30, 2010  Aug 20, 2013  Sas Institute Inc.  Systems and methods for multiechelon inventory planning with lateral transshipment 
US8639558  Sep 25, 2006  Jan 28, 2014  International Business Machines Corporation  Providing markdown item pricing and promotion calendar 
US8645191  Jan 10, 2011  Feb 4, 2014  Oracle International Corporation  Product pricing optimization system 
US8688497  Jan 10, 2011  Apr 1, 2014  Sas Institute Inc.  Systems and methods for determining pack allocations 
US8788315  Jan 10, 2011  Jul 22, 2014  Sas Institute Inc.  Systems and methods for determining pack allocations 
US8812338  Apr 29, 2008  Aug 19, 2014  Sas Institute Inc.  Computerimplemented systems and methods for pack optimization 
US20030110066 *  Dec 9, 2002  Jun 12, 2003  I2 Technologies Us, Inc.  Generating an optimized pricing plan 
US20030130883 *  Nov 22, 2002  Jul 10, 2003  Schroeder Glenn George  Business planner 
US20060161465 *  Dec 8, 2005  Jul 20, 2006  Ramakrishnan Vishwamitra S  Systems and methods for optimizing total merchandise profitability 
US20060173765 *  Mar 20, 2006  Aug 3, 2006  Level 3 Communications, Inc.  System and method for determining an evolving combination of network components to maximize the net present value of a provider's cash flow 
US20060218025 *  Mar 27, 2006  Sep 28, 2006  Miller John M  Variable pricing module 
US20080065435 *  Aug 25, 2006  Mar 13, 2008  John Phillip Ratzloff  Computerimplemented systems and methods for reducing cost flow models 
US20080077459 *  Sep 25, 2006  Mar 27, 2008  Demandtec Inc.  Price markdown apparatus 
US20080097886 *  Oct 18, 2006  Apr 24, 2008  Pricemetrix, Inc.  Reference price framework 
US20080262903 *  Jun 27, 2008  Oct 23, 2008  International Business Machines Corporation  System and method to determine the prices and order quantities that maximize a retailer's total profit 
US20080270221 *  Dec 18, 2007  Oct 30, 2008  Silvaris Corporation  Determining and presenting product market prices 
US20090083079 *  Sep 21, 2007  Mar 26, 2009  Sharon Dawn Law  System and method of processing a health insurance claim 
US20090119193 *  Nov 2, 2007  May 7, 2009  Selleck John C  Automated transaction calculations with scripted rule sets 
US20090144103 *  Nov 30, 2007  Jun 4, 2009  Sap Ag  Method and System Incorporating Business Rules and Price Optimization for Financial System 
US20090307152 *  Jun 9, 2008  Dec 10, 2009  Oracle International Corporation  Cost Management System 
US20110047095 *  Feb 24, 2011  David Charles Cushing  Systems and Methods for Order Analysis, Enrichment and Execution  
US20120259674 *  Apr 5, 2011  Oct 11, 2012  American Express Travel Related Services Company, Inc.  Systems and methods for providing market insight 
US20130006713 *  Jun 28, 2012  Jan 3, 2013  Derek Robert Haake  Method for aggregating pricing information and assigning a fair market value to goods sold in a peertopeer ecommerce transaction 
US20130013474 *  Jan 10, 2013  Charles Christopher Whelan  System and method for modifying an indexbased hierarchal cost model of a complex system  
US20130041721 *  Jan 4, 2011  Feb 14, 2013  Artnet Ag  Art evaluation engine and method for automatic development of an art index 
US20130275183 *  May 25, 2012  Oct 17, 2013  Oracle International Corporation  Timedependent product pricing optimizer 
US20130311239 *  Jul 22, 2013  Nov 21, 2013  Jda Software Group, Inc.  Generating an Optimized Pricing Plan 
US20140324546 *  Apr 25, 2013  Oct 30, 2014  International Business Machines Corporation  Simultaneous retail and transfer price optimizations 
US20150019297 *  Sep 27, 2014  Jan 15, 2015  International Business Machines Corporation  Simultaneous retail and transfer price optimizations 
US20160063520 *  Nov 9, 2015  Mar 3, 2016  Bank Of America Corporation  System for Analyzing Patterns 
U.S. Classification  705/7.35, 705/1.1, 705/37, 705/400 
International Classification  G06Q99/00 
Cooperative Classification  G06Q40/04, G06Q30/0283, G06Q30/06, G06Q30/0206 
European Classification  G06Q30/06, G06Q30/0283, G06Q30/0206, G06Q40/04 
Date  Code  Event  Description 

Feb 4, 2002  AS  Assignment  Owner name: SILICON VALLEY BANK,CALIFORNIA Free format text: SECURITY AGREEMENT;ASSIGNOR:DEMANDTEC, INC.;REEL/FRAME:012588/0668 Effective date: 20011023 Owner name: SILICON VALLEY BANK, CALIFORNIA Free format text: SECURITY AGREEMENT;ASSIGNOR:DEMANDTEC, INC.;REEL/FRAME:012588/0668 Effective date: 20011023 
Jul 2, 2003  AS  Assignment  Owner name: SILICON VALLEY BANK, CALIFORNIA Free format text: SECURITY INTEREST;ASSIGNOR:DEMANDTEC, INC.;REEL/FRAME:014216/0639 Effective date: 20030515 
Aug 9, 2006  AS  Assignment  Owner name: SILICON VALLEY BANK, CALIFORNIA Free format text: SECURITY AGREEMENT;ASSIGNOR:DEMANDTEC, INC.;REEL/FRAME:018420/0001 Effective date: 20060725 Owner name: SILICON VALLEY BANK,CALIFORNIA Free format text: SECURITY AGREEMENT;ASSIGNOR:DEMANDTEC, INC.;REEL/FRAME:018420/0001 Effective date: 20060725 
Dec 15, 2011  AS  Assignment  Owner name: DEMANDTEC INC., CALIFORNIA Free format text: RELEASE;ASSIGNOR:SILICON VALLEY BANK;REEL/FRAME:027419/0194 Effective date: 20111213 Owner name: DEMANDTEC INC., CALIFORNIA Free format text: RELEASE;ASSIGNOR:SILICON VALLEY BANK;REEL/FRAME:027419/0117 Effective date: 20111213 Owner name: DEMANDTEC INC., CALIFORNIA Free format text: RELEASE;ASSIGNOR:SILICON VALLEY BANK;REEL/FRAME:027419/0390 Effective date: 20111213 
Jan 10, 2013  AS  Assignment  Owner name: INTERNATIONAL BUSINESS MACHINES CORPORATION, NEW Y Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:DEMANDTEC INC.;REEL/FRAME:029604/0877 Effective date: 20130109 
Mar 29, 2013  FPAY  Fee payment  Year of fee payment: 4 